Bump debian requirement

* added language model as parameter

* fix fore more help explanation

CMakeLists.txt

@@ -0,0 +1,20 @@
+cmake_minimum_required(VERSION 3.13)
+project(vosk-api CXX)
+
+set(CMAKE_CXX_STANDARD 17)
+set(CMAKE_CXX_EXTENSIONS OFF)
+
+add_library(vosk
+  src/language_model.cc
+  src/model.cc
+  src/recognizer.cc
+  src/spk_model.cc
+  src/vosk_api.cc
+)
+
+find_package(kaldi REQUIRED)
+target_link_libraries(vosk PUBLIC kaldi-base kaldi-online2 kaldi-rnnlm fstngram)
+
+include(GNUInstallDirs)
+install(TARGETS vosk DESTINATION ${CMAKE_INSTALL_LIBDIR})
+install(FILES src/vosk_api.h DESTINATION ${CMAKE_INSTALL_INCLUDEDIR})

README.md

@@ -4,14 +4,15 @@ Vosk is an offline open source speech recognition toolkit. It enables
 speech recognition for 20+ languages and dialects - English, Indian
 English, German, French, Spanish, Portuguese, Chinese, Russian, Turkish,
 Vietnamese, Italian, Dutch, Catalan, Arabic, Greek, Farsi, Filipino,
-Ukrainian, Kazakh, Swedish, Japanese, Esperanto, Hindi, Czech. More to come.
+Ukrainian, Kazakh, Swedish, Japanese, Esperanto, Hindi, Czech, Polish.
+More to come.
 
 Vosk models are small (50 Mb) but provide continuous large vocabulary
 transcription, zero-latency response with streaming API, reconfigurable
 vocabulary and speaker identification.
 
 Speech recognition bindings implemented for various programming languages
-like Python, Java, Node.JS, C#, C++ and others.
+like Python, Java, Node.JS, C#, C++, Rust, Go and others.
 
 Vosk supplies speech recognition for chatbots, smart home appliances,
 virtual assistants. It can also create subtitles for movies,

android/build.gradle

@@ -10,7 +10,7 @@ buildscript {
 }
 
 allprojects {
-    version = '0.3.41'
+    version = '0.3.45'
 }
 
 subprojects {

csharp/demo/VoskDemo.csproj

@@ -11,7 +11,7 @@
   </PropertyGroup>
 
   <ItemGroup>
-    <PackageReference Include="Vosk" Version="0.3.41" />
+    <PackageReference Include="Vosk" Version="0.3.45" />
   </ItemGroup>
 
 </Project>

csharp/nuget/Vosk.nuspec

@@ -2,7 +2,7 @@
 <package>
   <metadata>
     <id>Vosk</id>
-    <version>0.3.41</version>
+    <version>0.3.45</version>
     <authors>Alpha Cephei Inc</authors>
     <owners>Alpha Cephei Inc</owners>
     <license type="expression">Apache-2.0</license>

go/README.md

@@ -0,0 +1 @@
+See example subfolder for instructions how to use the module

go/example/README.md

@@ -0,0 +1,31 @@
+To try this package do the following steps:
+
+On Linux (we download library and set LD_LIBRARY_PATH)
+
+```
+git clone https://github.com/alphacep/vosk-api
+cd vosk-api/go/example
+wget https://github.com/alphacep/vosk-api/releases/download/v0.3.45/vosk-linux-x86_64-0.3.45.zip
+unzip vosk-linux-x86_64-0.3.45.zip
+wget https://alphacephei.com/vosk/models/vosk-model-small-en-us-0.15.zip
+unzip vosk-model-small-en-us-0.15.zip
+mv vosk-model-small-en-us-0.15 model
+cp ../../python/example/test.wav .
+VOSK_PATH=`pwd`/vosk-linux-x86_64-0.3.45 LD_LIBRARY_PATH=$VOSK_PATH CGO_CPPFLAGS="-I $VOSK_PATH" CGO_LDFLAGS="-L $VOSK_PATH" go run . -f test.wav
+```
+
+for Windows (we place DLLs in current folder where linker finds them):
+
+```
+git clone https://github.com/alphacep/vosk-api
+cd vosk-api/go/example
+wget https://github.com/alphacep/vosk-api/releases/download/v0.3.45/vosk-linux-x86_64-0.3.45.zip
+unzip vosk-linux-x86_64-0.3.45.zip
+cp vosk-linux-x86_64-0.3.45/*.dll .
+cp vosk-linux-x86_64-0.3.45/*.h .
+wget https://alphacephei.com/vosk/models/vosk-model-small-en-us-0.15.zip
+unzip vosk-model-small-en-us-0.15.zip
+mv vosk-model-small-en-us-0.15 model
+cp ../../python/example/test.wav .
+VOSK_PATH=`pwd` LD_LIBRARY_PATH=$VOSK_PATH CGO_CPPFLAGS="-I $VOSK_PATH" CGO_LDFLAGS="-L $VOSK_PATH -lvosk -lpthread -dl" go run . -f test.wav
+```

go/example/doc.go

@@ -1,2 +1,2 @@
 // Example package for Vosk Go bindings.
-package test_simple
+package main

go/example/test_simple.go

@@ -7,6 +7,7 @@ import (
 	"io"
 	"log"
 	"os"
+	"encoding/json"
 
 	vosk "github.com/alphacep/vosk-api/go"
 )
@@ -21,6 +22,9 @@ func main() {
 		log.Fatal(err)
 	}
 
+	// we can check if word is in the vocabulary
+	// fmt.Println(model.FindWord("air"))
+
 	sampleRate := 16000.0
 	rec, err := vosk.NewRecognizer(model, sampleRate)
 	if err != nil {
@@ -48,9 +52,12 @@ func main() {
 		}
 
 		if rec.AcceptWaveform(buf) != 0 {
-			fmt.Println(string(rec.Result()))
+			fmt.Println(rec.Result())
 		}
 	}
 
-	fmt.Println(string(rec.FinalResult()))
+	// Unmarshal example for final result
+	var jres map[string]interface{}
+	json.Unmarshal([]byte(rec.FinalResult()), &jres)
+	fmt.Println(jres["text"])
 }

go/vosk.go

@@ -5,6 +5,7 @@ package vosk
 // #include <stdlib.h>
 // #include <vosk_api.h>
 import "C"
+import "unsafe"
 
 // VoskModel contains a reference to the C VoskModel
 type VoskModel struct {
@@ -13,7 +14,9 @@ type VoskModel struct {
 
 // NewModel creates a new VoskModel instance
 func NewModel(modelPath string) (*VoskModel, error) {
-	internal := C.vosk_model_new(C.CString(modelPath))
+	cmodelPath := C.CString(modelPath)
+	defer C.free(unsafe.Pointer(cmodelPath))
+	internal := C.vosk_model_new(cmodelPath)
 	model := &VoskModel{model: internal}
 	return model, nil
 }
@@ -29,10 +32,10 @@ func freeModel(model *VoskModel) {
 // FindWord checks if a word can be recognized by the model.
 // Returns the word symbol if the word exists inside the model or
 // -1 otherwise.
-func (m *VoskModel) FindWord(word []byte) int {
-	cbuf := C.CBytes(word)
-	defer C.free(cbuf)
-	i := C.vosk_model_find_word(m.model, (*C.char)(cbuf))
+func (m *VoskModel) FindWord(word string) int {
+	cstr := C.CString(word)
+	defer C.free(unsafe.Pointer(cstr))
+	i := C.vosk_model_find_word(m.model, cstr)
 	return int(i)
 }
 
@@ -43,7 +46,9 @@ type VoskSpkModel struct {
 
 // NewSpkModel creates a new VoskSpkModel instance
 func NewSpkModel(spkModelPath string) (*VoskSpkModel, error) {
-	internal := C.vosk_spk_model_new(C.CString(spkModelPath))
+	cspkModelPath := C.CString(spkModelPath)
+	defer C.free(unsafe.Pointer(cspkModelPath))
+	internal := C.vosk_spk_model_new(cspkModelPath)
 	spkModel := &VoskSpkModel{spkModel: internal}
 	return spkModel, nil
 }
@@ -84,10 +89,10 @@ func NewRecognizerSpk(model *VoskModel, sampleRate float64, spkModel *VoskSpkMod
 }
 
 // NewRecognizerGrm creates a new VoskRecognizer instance with the phrase list.
-func NewRecognizerGrm(model *VoskModel, sampleRate float64, grammer []byte) (*VoskRecognizer, error) {
-	cbuf := C.CBytes(grammer)
-	defer C.free(cbuf)
-	internal := C.vosk_recognizer_new_grm(model.model, C.float(sampleRate), (*C.char)(cbuf))
+func NewRecognizerGrm(model *VoskModel, sampleRate float64, grammar string) (*VoskRecognizer, error) {
+	cgrammar := C.CString(grammar)
+	defer C.free(unsafe.Pointer(cgrammar))
+	internal := C.vosk_recognizer_new_grm(model.model, C.float(sampleRate), cgrammar)
 	rec := &VoskRecognizer{rec: internal}
 	return rec, nil
 }
@@ -107,6 +112,11 @@ func (r *VoskRecognizer) SetWords(words int) {
 	C.vosk_recognizer_set_words(r.rec, C.int(words))
 }
 
+// SetPartialWords enables words with times in the partial ouput.
+func (r *VoskRecognizer) SetPartialWords(words int) {
+	C.vosk_recognizer_set_partial_words(r.rec, C.int(words))
+}
+
 // AcceptWaveform accepts and processes a new chunk of the voice data.
 func (r *VoskRecognizer) AcceptWaveform(buffer []byte) int {
 	cbuf := C.CBytes(buffer)
@@ -116,19 +126,19 @@ func (r *VoskRecognizer) AcceptWaveform(buffer []byte) int {
 }
 
 // Result returns a speech recognition result.
-func (r *VoskRecognizer) Result() []byte {
-	return []byte(C.GoString(C.vosk_recognizer_result(r.rec)))
+func (r *VoskRecognizer) Result() string {
+	return C.GoString(C.vosk_recognizer_result(r.rec))
 }
 
 // PartialResult returns a partial speech recognition result.
-func (r *VoskRecognizer) PartialResult() []byte {
-	return []byte(C.GoString(C.vosk_recognizer_partial_result(r.rec)))
+func (r *VoskRecognizer) PartialResult() string {
+	return C.GoString(C.vosk_recognizer_partial_result(r.rec))
 }
 
 // FinalResult returns a speech recognition result. Same as result, but doesn't wait
 // for silence.
-func (r *VoskRecognizer) FinalResult() []byte {
-	return []byte(C.GoString(C.vosk_recognizer_final_result(r.rec)))
+func (r *VoskRecognizer) FinalResult() string {
+	return C.GoString(C.vosk_recognizer_final_result(r.rec))
 }
 
 // Reset resets the recognizer.

java/demo/build.gradle

@@ -12,5 +12,5 @@ repositories {
 
 dependencies {
     implementation group: 'net.java.dev.jna', name: 'jna', version: '5.7.0'
-    implementation group: 'com.alphacephei', name: 'vosk', version: '0.3.41+'
+    implementation group: 'com.alphacephei', name: 'vosk', version: '0.3.40+'
 }

java/lib/build.gradle

@@ -16,7 +16,7 @@ repositories {
 
 archivesBaseName = 'vosk'
 group = 'com.alphacephei'
-version = '0.3.41'
+version = '0.3.45'
 
 mavenPublish {
     group = 'com.alphacephei'

nodejs/README.md

@@ -13,7 +13,8 @@ Vosk is an offline open source speech recognition toolkit. It enables
 speech recognition for 20+ languages and dialects - English, Indian
 English, German, French, Spanish, Portuguese, Chinese, Russian, Turkish,
 Vietnamese, Italian, Dutch, Catalan, Arabic, Greek, Farsi, Filipino,
-Ukrainian, Kazakh, Swedish, Japanese, Esperanto, Hindi, Czech. More to come.
+Ukrainian, Kazakh, Swedish, Japanese, Esperanto, Hindi, Czech, Polish.
+More to come.
 
 Vosk models are small (50 Mb) but provide continuous large vocabulary
 transcription, zero-latency response with streaming API, reconfigurable

nodejs/demo/test_microphone.js

@@ -18,11 +18,11 @@ const rec = new vosk.Recognizer({model: model, sampleRate: SAMPLE_RATE});
 var micInstance = mic({
     rate: String(SAMPLE_RATE),
     channels: '1',
-    debug: false
+    debug: false,
+    device: 'default',    
 });
 
 var micInputStream = micInstance.getAudioStream();
-micInstance.start();
 
 micInputStream.on('data', data => {
     if (rec.acceptWaveform(data))
@@ -31,9 +31,16 @@ micInputStream.on('data', data => {
         console.log(rec.partialResult());
 });
 
-process.on('SIGINT', function() {
+micInputStream.on('audioProcessExitComplete', function() {
+    console.log("Cleaning up");
     console.log(rec.finalResult());
-    console.log("\nDone");
     rec.free();
     model.free();
 });
+
+process.on('SIGINT', function() {
+    console.log("\nStopping");
+    micInstance.stop();
+});
+
+micInstance.start();

nodejs/index.js

@@ -72,7 +72,7 @@ if (os.platform() == 'win32') {
     // Update path to load dependent dlls
     let currentPath = process.env.Path;
     let dllDirectory = path.resolve(path.join(__dirname, "lib", "win-x86_64"));
-    process.env.Path = currentPath + path.delimiter + dllDirectory;
+    process.env.Path = dllDirectory + path.delimiter + currentPath;
 
     soname = path.join(__dirname, "lib", "win-x86_64", "libvosk.dll")
 } else if (os.platform() == 'darwin') {

nodejs/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vosk",
-  "version": "0.3.41",
+  "version": "0.3.45",
   "description": "Node binding for continuous offline voice recoginition with Vosk library.",
   "repository": {
     "type": "git",

python/.pylintrc

@@ -0,0 +1,639 @@
+[MAIN]
+
+# Analyse import fallback blocks. This can be used to support both Python 2 and
+# 3 compatible code, which means that the block might have code that exists
+# only in one or another interpreter, leading to false positives when analysed.
+analyse-fallback-blocks=no
+
+# Load and enable all available extensions. Use --list-extensions to see a list
+# all available extensions.
+#enable-all-extensions=
+
+# In error mode, messages with a category besides ERROR or FATAL are
+# suppressed, and no reports are done by default. Error mode is compatible with
+# disabling specific errors.
+#errors-only=
+
+# Always return a 0 (non-error) status code, even if lint errors are found.
+# This is primarily useful in continuous integration scripts.
+#exit-zero=
+
+# A comma-separated list of package or module names from where C extensions may
+# be loaded. Extensions are loading into the active Python interpreter and may
+# run arbitrary code.
+extension-pkg-allow-list=
+
+# A comma-separated list of package or module names from where C extensions may
+# be loaded. Extensions are loading into the active Python interpreter and may
+# run arbitrary code. (This is an alternative name to extension-pkg-allow-list
+# for backward compatibility.)
+extension-pkg-whitelist=
+
+# Return non-zero exit code if any of these messages/categories are detected,
+# even if score is above --fail-under value. Syntax same as enable. Messages
+# specified are enabled, while categories only check already-enabled messages.
+fail-on=
+
+# Specify a score threshold under which the program will exit with error.
+fail-under=10
+
+# Interpret the stdin as a python script, whose filename needs to be passed as
+# the module_or_package argument.
+#from-stdin=
+
+# Files or directories to be skipped. They should be base names, not paths.
+ignore=CVS
+
+# Add files or directories matching the regular expressions patterns to the
+# ignore-list. The regex matches against paths and can be in Posix or Windows
+# format. Because '\' represents the directory delimiter on Windows systems, it
+# can't be used as an escape character.
+ignore-paths=
+
+# Files or directories matching the regular expression patterns are skipped.
+# The regex matches against base names, not paths. The default value ignores
+# Emacs file locks
+ignore-patterns=^\.#
+
+# List of module names for which member attributes should not be checked
+# (useful for modules/projects where namespaces are manipulated during runtime
+# and thus existing member attributes cannot be deduced by static analysis). It
+# supports qualified module names, as well as Unix pattern matching.
+ignored-modules=
+
+# Python code to execute, usually for sys.path manipulation such as
+# pygtk.require().
+#init-hook=
+
+# Use multiple processes to speed up Pylint. Specifying 0 will auto-detect the
+# number of processors available to use, and will cap the count on Windows to
+# avoid hangs.
+jobs=1
+
+# Control the amount of potential inferred values when inferring a single
+# object. This can help the performance when dealing with large functions or
+# complex, nested conditions.
+limit-inference-results=100
+
+# List of plugins (as comma separated values of python module names) to load,
+# usually to register additional checkers.
+load-plugins=
+
+# Pickle collected data for later comparisons.
+persistent=yes
+
+# Minimum Python version to use for version dependent checks. Will default to
+# the version used to run pylint.
+py-version=3.7
+
+# Discover python modules and packages in the file system subtree.
+recursive=no
+
+# When enabled, pylint would attempt to guess common misconfiguration and emit
+# user-friendly hints instead of false-positive error messages.
+suggestion-mode=yes
+
+# Allow loading of arbitrary C extensions. Extensions are imported into the
+# active Python interpreter and may run arbitrary code.
+unsafe-load-any-extension=no
+
+# In verbose mode, extra non-checker-related info will be displayed.
+#verbose=
+
+
+[REPORTS]
+
+# Python expression which should return a score less than or equal to 10. You
+# have access to the variables 'fatal', 'error', 'warning', 'refactor',
+# 'convention', and 'info' which contain the number of messages in each
+# category, as well as 'statement' which is the total number of statements
+# analyzed. This score is used by the global evaluation report (RP0004).
+evaluation=max(0, 0 if fatal else 10.0 - ((float(5 * error + warning + refactor + convention) / statement) * 10))
+
+# Template used to display messages. This is a python new-style format string
+# used to format the message information. See doc for all details.
+msg-template=
+
+# Set the output format. Available formats are text, parseable, colorized, json
+# and msvs (visual studio). You can also give a reporter class, e.g.
+# mypackage.mymodule.MyReporterClass.
+#output-format=
+
+# Tells whether to display a full report or only the messages.
+reports=no
+
+# Activate the evaluation score.
+score=yes
+
+
+[MESSAGES CONTROL]
+
+# Only show warnings with the listed confidence levels. Leave empty to show
+# all. Valid levels: HIGH, CONTROL_FLOW, INFERENCE, INFERENCE_FAILURE,
+# UNDEFINED.
+confidence=HIGH,
+           CONTROL_FLOW,
+           INFERENCE,
+           INFERENCE_FAILURE,
+           UNDEFINED
+
+# Disable the message, report, category or checker with the given id(s). You
+# can either give multiple identifiers separated by comma (,) or put this
+# option multiple times (only on the command line, not in the configuration
+# file where it should appear only once). You can also use "--disable=all" to
+# disable everything first and then re-enable specific checks. For example, if
+# you want to run only the similarities checker, you can use "--disable=all
+# --enable=similarities". If you want to run only the classes checker, but have
+# no Warning level messages displayed, use "--disable=all --enable=classes
+# --disable=W".
+disable=raw-checker-failed,
+        bad-inline-option,
+        locally-disabled,
+        file-ignored,
+        suppressed-message,
+        useless-suppression,
+        deprecated-pragma,
+        use-symbolic-message-instead,
+	missing-module-docstring, # added
+	missing-class-docstring, # added
+	missing-function-docstring, # added
+	invalid-name, # added
+	broad-except, # added
+	wrong-import-order, # added
+	consider-using-f-string, # added
+	logging-format-interpolation, # added
+	no-member, # added
+	consider-using-with, # added
+	unused-argument,
+	no-else-return, # added
+	too-few-public-methods, # added
+	import-error, # added
+	protected-access, # added
+	unused-variable, # added
+	R0801 # added (similarity of two scripts)
+
+# Enable the message, report, category or checker with the given id(s). You can
+# either give multiple identifier separated by comma (,) or put this option
+# multiple time (only on the command line, not in the configuration file where
+# it should appear only once). See also the "--disable" option for examples.
+enable=c-extension-no-member
+
+
+[BASIC]
+
+# Naming style matching correct argument names.
+argument-naming-style=snake_case
+
+# Regular expression matching correct argument names. Overrides argument-
+# naming-style. If left empty, argument names will be checked with the set
+# naming style.
+#argument-rgx=
+
+# Naming style matching correct attribute names.
+attr-naming-style=snake_case
+
+# Regular expression matching correct attribute names. Overrides attr-naming-
+# style. If left empty, attribute names will be checked with the set naming
+# style.
+#attr-rgx=
+
+# Bad variable names which should always be refused, separated by a comma.
+bad-names=foo,
+          bar,
+          baz,
+          toto,
+          tutu,
+          tata
+
+# Bad variable names regexes, separated by a comma. If names match any regex,
+# they will always be refused
+bad-names-rgxs=
+
+# Naming style matching correct class attribute names.
+class-attribute-naming-style=any
+
+# Regular expression matching correct class attribute names. Overrides class-
+# attribute-naming-style. If left empty, class attribute names will be checked
+# with the set naming style.
+#class-attribute-rgx=
+
+# Naming style matching correct class constant names.
+class-const-naming-style=UPPER_CASE
+
+# Regular expression matching correct class constant names. Overrides class-
+# const-naming-style. If left empty, class constant names will be checked with
+# the set naming style.
+#class-const-rgx=
+
+# Naming style matching correct class names.
+class-naming-style=PascalCase
+
+# Regular expression matching correct class names. Overrides class-naming-
+# style. If left empty, class names will be checked with the set naming style.
+#class-rgx=
+
+# Naming style matching correct constant names.
+const-naming-style=UPPER_CASE
+
+# Regular expression matching correct constant names. Overrides const-naming-
+# style. If left empty, constant names will be checked with the set naming
+# style.
+#const-rgx=
+
+# Minimum line length for functions/classes that require docstrings, shorter
+# ones are exempt.
+docstring-min-length=-1
+
+# Naming style matching correct function names.
+function-naming-style=snake_case
+
+# Regular expression matching correct function names. Overrides function-
+# naming-style. If left empty, function names will be checked with the set
+# naming style.
+#function-rgx=
+
+# Good variable names which should always be accepted, separated by a comma.
+good-names=i,
+           j,
+           k,
+           ex,
+           Run,
+           _,
+	   x,
+	   y,
+	   nx,
+	   ny
+
+# Good variable names regexes, separated by a comma. If names match any regex,
+# they will always be accepted
+good-names-rgxs=
+
+# Include a hint for the correct naming format with invalid-name.
+include-naming-hint=no
+
+# Naming style matching correct inline iteration names.
+inlinevar-naming-style=any
+
+# Regular expression matching correct inline iteration names. Overrides
+# inlinevar-naming-style. If left empty, inline iteration names will be checked
+# with the set naming style.
+#inlinevar-rgx=
+
+# Naming style matching correct method names.
+method-naming-style=snake_case
+
+# Regular expression matching correct method names. Overrides method-naming-
+# style. If left empty, method names will be checked with the set naming style.
+#method-rgx=
+
+# Naming style matching correct module names.
+module-naming-style=snake_case
+
+# Regular expression matching correct module names. Overrides module-naming-
+# style. If left empty, module names will be checked with the set naming style.
+#module-rgx=
+
+# Colon-delimited sets of names that determine each other's naming style when
+# the name regexes allow several styles.
+name-group=
+
+# Regular expression which should only match function or class names that do
+# not require a docstring.
+no-docstring-rgx=^_
+
+# List of decorators that produce properties, such as abc.abstractproperty. Add
+# to this list to register other decorators that produce valid properties.
+# These decorators are taken in consideration only for invalid-name.
+property-classes=abc.abstractproperty
+
+# Regular expression matching correct type variable names. If left empty, type
+# variable names will be checked with the set naming style.
+#typevar-rgx=
+
+# Naming style matching correct variable names.
+variable-naming-style=snake_case
+
+# Regular expression matching correct variable names. Overrides variable-
+# naming-style. If left empty, variable names will be checked with the set
+# naming style.
+#variable-rgx=
+
+
+[VARIABLES]
+
+# List of additional names supposed to be defined in builtins. Remember that
+# you should avoid defining new builtins when possible.
+additional-builtins=
+
+# Tells whether unused global variables should be treated as a violation.
+allow-global-unused-variables=yes
+
+# List of names allowed to shadow builtins
+allowed-redefined-builtins=
+
+# List of strings which can identify a callback function by name. A callback
+# name must start or end with one of those strings.
+callbacks=cb_,
+          _cb
+
+# A regular expression matching the name of dummy variables (i.e. expected to
+# not be used).
+dummy-variables-rgx=_+$|(_[a-zA-Z0-9_]*[a-zA-Z0-9]+?$)|dummy|^ignored_|^unused_
+
+# Argument names that match this expression will be ignored.
+ignored-argument-names=_.*|^ignored_|^unused_
+
+# Tells whether we should check for unused import in __init__ files.
+init-import=no
+
+# List of qualified module names which can have objects that can redefine
+# builtins.
+redefining-builtins-modules=six.moves,past.builtins,future.builtins,builtins,io
+
+
+[SIMILARITIES]
+
+# Comments are removed from the similarity computation
+ignore-comments=yes
+
+# Docstrings are removed from the similarity computation
+ignore-docstrings=yes
+
+# Imports are removed from the similarity computation
+ignore-imports=yes
+
+# Signatures are removed from the similarity computation
+ignore-signatures=yes
+
+# Minimum lines number of a similarity.
+min-similarity-lines=4
+
+
+[MISCELLANEOUS]
+
+# List of note tags to take in consideration, separated by a comma.
+notes=FIXME,
+      XXX,
+      TODO
+
+# Regular expression of note tags to take in consideration.
+notes-rgx=
+
+
+[STRING]
+
+# This flag controls whether inconsistent-quotes generates a warning when the
+# character used as a quote delimiter is used inconsistently within a module.
+check-quote-consistency=no
+
+# This flag controls whether the implicit-str-concat should generate a warning
+# on implicit string concatenation in sequences defined over several lines.
+check-str-concat-over-line-jumps=no
+
+
+[IMPORTS]
+
+# List of modules that can be imported at any level, not just the top level
+# one.
+allow-any-import-level=
+
+# Allow wildcard imports from modules that define __all__.
+allow-wildcard-with-all=no
+
+# Deprecated modules which should not be used, separated by a comma.
+deprecated-modules=
+
+# Output a graph (.gv or any supported image format) of external dependencies
+# to the given file (report RP0402 must not be disabled).
+ext-import-graph=
+
+# Output a graph (.gv or any supported image format) of all (i.e. internal and
+# external) dependencies to the given file (report RP0402 must not be
+# disabled).
+import-graph=
+
+# Output a graph (.gv or any supported image format) of internal dependencies
+# to the given file (report RP0402 must not be disabled).
+int-import-graph=
+
+# Force import order to recognize a module as part of the standard
+# compatibility libraries.
+known-standard-library=
+
+# Force import order to recognize a module as part of a third party library.
+known-third-party=enchant
+
+# Couples of modules and preferred modules, separated by a comma.
+preferred-modules=
+
+
+[EXCEPTIONS]
+
+# Exceptions that will emit a warning when caught.
+overgeneral-exceptions=BaseException,
+                       Exception
+
+
+[DESIGN]
+
+# List of regular expressions of class ancestor names to ignore when counting
+# public methods (see R0903)
+exclude-too-few-public-methods=
+
+# List of qualified class names to ignore when counting class parents (see
+# R0901)
+ignored-parents=
+
+# Maximum number of arguments for function / method.
+max-args=5
+
+# Maximum number of attributes for a class (see R0902).
+max-attributes=7
+
+# Maximum number of boolean expressions in an if statement (see R0916).
+max-bool-expr=5
+
+# Maximum number of branch for function / method body.
+max-branches=12
+
+# Maximum number of locals for function / method body.
+max-locals=15
+
+# Maximum number of parents for a class (see R0901).
+max-parents=7
+
+# Maximum number of public methods for a class (see R0904).
+max-public-methods=20
+
+# Maximum number of return / yield for function / method body.
+max-returns=6
+
+# Maximum number of statements in function / method body.
+max-statements=50
+
+# Minimum number of public methods for a class (see R0903).
+min-public-methods=2
+
+
+[METHOD_ARGS]
+
+# List of qualified names (i.e., library.method) which require a timeout
+# parameter e.g. 'requests.api.get,requests.api.post'
+timeout-methods=requests.api.delete,requests.api.get,requests.api.head,requests.api.options,requests.api.patch,requests.api.post,requests.api.put,requests.api.request
+
+
+[LOGGING]
+
+# The type of string formatting that logging methods do. `old` means using %
+# formatting, `new` is for `{}` formatting.
+logging-format-style=old
+
+# Logging modules to check that the string format arguments are in logging
+# function parameter format.
+logging-modules=logging
+
+
+[FORMAT]
+
+# Expected format of line ending, e.g. empty (any line ending), LF or CRLF.
+expected-line-ending-format=
+
+# Regexp for a line that is allowed to be longer than the limit.
+ignore-long-lines=^\s*(# )?<?https?://\S+>?$
+
+# Number of spaces of indent required inside a hanging or continued line.
+indent-after-paren=4
+
+# String used as indentation unit. This is usually "    " (4 spaces) or "\t" (1
+# tab).
+indent-string='    '
+
+# Maximum number of characters on a single line.
+max-line-length=100
+
+# Maximum number of lines in a module.
+max-module-lines=1000
+
+# Allow the body of a class to be on the same line as the declaration if body
+# contains single statement.
+single-line-class-stmt=no
+
+# Allow the body of an if to be on the same line as the test if there is no
+# else.
+single-line-if-stmt=no
+
+
+[REFACTORING]
+
+# Maximum number of nested blocks for function / method body
+max-nested-blocks=5
+
+# Complete name of functions that never returns. When checking for
+# inconsistent-return-statements if a never returning function is called then
+# it will be considered as an explicit return statement and no message will be
+# printed.
+never-returning-functions=sys.exit,argparse.parse_error
+
+
+[TYPECHECK]
+
+# List of decorators that produce context managers, such as
+# contextlib.contextmanager. Add to this list to register other decorators that
+# produce valid context managers.
+contextmanager-decorators=contextlib.contextmanager
+
+# List of members which are set dynamically and missed by pylint inference
+# system, and so shouldn't trigger E1101 when accessed. Python regular
+# expressions are accepted.
+generated-members=
+
+# Tells whether to warn about missing members when the owner of the attribute
+# is inferred to be None.
+ignore-none=yes
+
+# This flag controls whether pylint should warn about no-member and similar
+# checks whenever an opaque object is returned when inferring. The inference
+# can return multiple potential results while evaluating a Python object, but
+# some branches might not be evaluated, which results in partial inference. In
+# that case, it might be useful to still emit no-member and other checks for
+# the rest of the inferred objects.
+ignore-on-opaque-inference=yes
+
+# List of symbolic message names to ignore for Mixin members.
+ignored-checks-for-mixins=no-member,
+                          not-async-context-manager,
+                          not-context-manager,
+                          attribute-defined-outside-init
+
+# List of class names for which member attributes should not be checked (useful
+# for classes with dynamically set attributes). This supports the use of
+# qualified names.
+ignored-classes=optparse.Values,thread._local,_thread._local,argparse.Namespace
+
+# Show a hint with possible names when a member name was not found. The aspect
+# of finding the hint is based on edit distance.
+missing-member-hint=yes
+
+# The minimum edit distance a name should have in order to be considered a
+# similar match for a missing member name.
+missing-member-hint-distance=1
+
+# The total number of similar names that should be taken in consideration when
+# showing a hint for a missing member.
+missing-member-max-choices=1
+
+# Regex pattern to define which classes are considered mixins.
+mixin-class-rgx=.*[Mm]ixin
+
+# List of decorators that change the signature of a decorated function.
+signature-mutators=
+
+
+[CLASSES]
+
+# Warn about protected attribute access inside special methods
+check-protected-access-in-special-methods=no
+
+# List of method names used to declare (i.e. assign) instance attributes.
+defining-attr-methods=__init__,
+                      __new__,
+                      setUp,
+                      __post_init__
+
+# List of member names, which should be excluded from the protected access
+# warning.
+exclude-protected=_asdict,
+                  _fields,
+                  _replace,
+                  _source,
+                  _make
+
+# List of valid names for the first argument in a class method.
+valid-classmethod-first-arg=cls
+
+# List of valid names for the first argument in a metaclass class method.
+valid-metaclass-classmethod-first-arg=cls
+
+
+[SPELLING]
+
+# Limits count of emitted suggestions for spelling mistakes.
+max-spelling-suggestions=4
+
+# Spelling dictionary name. Available dictionaries: none. To make it work,
+# install the 'python-enchant' package.
+spelling-dict=
+
+# List of comma separated words that should be considered directives if they
+# appear at the beginning of a comment and should not be checked.
+spelling-ignore-comment-directives=fmt: on,fmt: off,noqa:,noqa,nosec,isort:skip,mypy:
+
+# List of comma separated words that should not be checked.
+spelling-ignore-words=
+
+# A path to a file that contains the private dictionary; one word per line.
+spelling-private-dict-file=
+
+# Tells whether to store unknown words to the private dictionary (see the
+# --spelling-private-dict-file option) instead of raising a message.
+spelling-store-unknown-words=no

python/README.md

@@ -4,7 +4,8 @@ Vosk is an offline open source speech recognition toolkit. It enables
 speech recognition for 20+ languages and dialects - English, Indian
 English, German, French, Spanish, Portuguese, Chinese, Russian, Turkish,
 Vietnamese, Italian, Dutch, Catalan, Arabic, Greek, Farsi, Filipino,
-Ukrainian, Kazakh, Swedish, Japanese, Esperanto, Hindi, Czech. More to come.
+Ukrainian, Kazakh, Swedish, Japanese, Esperanto, Hindi, Czech, Polish.
+More to come.
 
 Vosk models are small (50 Mb) but provide continuous large vocabulary
 transcription, zero-latency response with streaming API, reconfigurable

python/example/colab/vosk-adaptation.ipynb

@@ -0,0 +1,558 @@
+{
+  "nbformat": 4,
+  "nbformat_minor": 0,
+  "metadata": {
+    "colab": {
+      "name": "Vosk Adaptation",
+      "provenance": [],
+      "collapsed_sections": []
+    },
+    "kernelspec": {
+      "name": "python3",
+      "display_name": "Python 3"
+    },
+    "language_info": {
+      "name": "python"
+    },
+    "accelerator": "GPU",
+    "gpuClass": "standard"
+  },
+  "cells": [
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {
+        "colab": {
+          "base_uri": "https://localhost:8080/"
+        },
+        "id": "URzWMmv50-Ba",
+        "outputId": "0e096a99-74dd-42e2-efb1-9cba784c3664"
+      },
+      "outputs": [
+        {
+          "output_type": "stream",
+          "name": "stdout",
+          "text": [
+            "/content\n",
+            "--2022-08-17 09:48:52--  https://alphacephei.com/vosk-colab/kaldi.tar.gz\n",
+            "Resolving alphacephei.com (alphacephei.com)... 188.40.21.16, 2a01:4f8:13a:279f::2\n",
+            "Connecting to alphacephei.com (alphacephei.com)|188.40.21.16|:443... connected.\n",
+            "HTTP request sent, awaiting response... 200 OK\n",
+            "Length: 809174554 (772M) [application/octet-stream]\n",
+            "Saving to: ‘kaldi.tar.gz’\n",
+            "\n",
+            "kaldi.tar.gz        100%[===================>] 771.69M  20.3MB/s    in 40s     \n",
+            "\n",
+            "2022-08-17 09:49:33 (19.4 MB/s) - ‘kaldi.tar.gz’ saved [809174554/809174554]\n",
+            "\n"
+          ]
+        }
+      ],
+      "source": [
+        "%cd /content\n",
+        "!wget -c https://alphacephei.com/vosk-colab/kaldi.tar.gz\n",
+        "!tar xzf kaldi.tar.gz"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "%cd /content/kaldi/egs/ac\n",
+        "!wget -c https://alphacephei.com/vosk-colab/vosk-model-small-en-us-0.15-compile-colab.tar.gz\n",
+        "!rm -rf vosk-model-small-en-us-0.15-compile-colab\n",
+        "!tar xf vosk-model-small-en-us-0.15-compile-colab.tar.gz"
+      ],
+      "metadata": {
+        "id": "-065p7WC2SHh",
+        "colab": {
+          "base_uri": "https://localhost:8080/"
+        },
+        "outputId": "241c7473-7464-48d5-b48d-dc6e3bf4971d"
+      },
+      "execution_count": 8,
+      "outputs": [
+        {
+          "output_type": "stream",
+          "name": "stdout",
+          "text": [
+            "/content/kaldi/egs/ac\n",
+            "--2022-08-17 10:28:26--  https://alphacephei.com/vosk-colab/vosk-model-small-en-us-0.15-compile-colab.tar.gz\n",
+            "Resolving alphacephei.com (alphacephei.com)... 188.40.21.16, 2a01:4f8:13a:279f::2\n",
+            "Connecting to alphacephei.com (alphacephei.com)|188.40.21.16|:443... connected.\n",
+            "HTTP request sent, awaiting response... 200 OK\n",
+            "Length: 59618100 (57M) [application/octet-stream]\n",
+            "Saving to: ‘vosk-model-small-en-us-0.15-compile-colab.tar.gz’\n",
+            "\n",
+            "vosk-model-small-en 100%[===================>]  56.86M  18.6MB/s    in 3.6s    \n",
+            "\n",
+            "2022-08-17 10:28:30 (15.7 MB/s) - ‘vosk-model-small-en-us-0.15-compile-colab.tar.gz’ saved [59618100/59618100]\n",
+            "\n"
+          ]
+        }
+      ]
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "%cd /content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab\n",
+        "!ls\n",
+        "!cat compile-graph.sh\n",
+        "!bash compile-graph.sh"
+      ],
+      "metadata": {
+        "colab": {
+          "base_uri": "https://localhost:8080/"
+        },
+        "id": "wuDjvNbd2sf9",
+        "outputId": "34a1d2fe-d443-4574-e25d-824e38eb3a78"
+      },
+      "execution_count": 9,
+      "outputs": [
+        {
+          "output_type": "stream",
+          "name": "stdout",
+          "text": [
+            "/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab\n",
+            "compile-graph.sh  data_test  decode.sh\texp\t      local  path.sh  steps\n",
+            "conf\t\t  db\t     dict.py\tget_vocab.py  mfcc   RESULTS  utils\n",
+            "#!/bin/bash\n",
+            "\n",
+            "set -x\n",
+            "\n",
+            ". path.sh\n",
+            "\n",
+            "pip3 install phonetisaurus\n",
+            "\n",
+            "rm -rf data\n",
+            "rm -rf exp/tdnn/lgraph\n",
+            "rm -rf exp/tdnn/lgraph_orig\n",
+            "\n",
+            "mkdir -p data/dict\n",
+            "cp db/phone/* data/dict\n",
+            "./dict.py > data/dict/lexicon.txt\n",
+            "\n",
+            "python3 ./get_vocab.py > data/mix.vocab\n",
+            "ngramsymbols data/mix.vocab data/mix.syms\n",
+            "farcompilestrings --fst_type=compact --symbols=data/mix.syms --keep_symbols --unknown_symbol=\"[unk]\" db/extra.txt data/extra.far\n",
+            "ngramcount --order=3 data/extra.far - |\n",
+            "    ngramprint --integers | grep -v \"<unk>\" | ngramread |\n",
+            "    ngramshrink --method=count_prune --count_pattern=\"3+:3\" |\n",
+            "    ngrammake --method=witten_bell - data/extra.mod\n",
+            "gunzip -c db/en-50k-0.4-android.lm.gz | ngramread --renormalize_arpa --ARPA --symbols=data/mix.syms - data/en-us.mod\n",
+            "ngrammerge --method=\"bayes_model_merge\" --normalize --alpha=0.95 --beta=0.05 data/en-us.mod data/extra.mod data/en-us-mix.mod\n",
+            "ngramprint --ARPA data/en-us-mix.mod | gzip -c > data/en-us-mix.lm.gz\n",
+            "\n",
+            "# Prune for the first stage if needed\n",
+            "# ngramshrink --method=relative_entropy --theta=2e-8 data/en-us-mix.mod data/en-us-mix-prune.mod\n",
+            "# ngramprint --ARPA data/en-us-mix-prune.mod | gzip -c > data/en-us-mix-small.lm.gz\n",
+            "\n",
+            "utils/prepare_lang.sh data/dict \"[unk]\" data/lang_local data/lang\n",
+            "utils/format_lm.sh data/lang db/en-50k-0.4-android.lm.gz data/dict/lexicon.txt data/lang_test\n",
+            "utils/format_lm.sh data/lang data/en-us-mix.lm.gz data/dict/lexicon.txt data/lang_test_adapt\n",
+            "\n",
+            "utils/mkgraph.sh --self-loop-scale 1.0 data/lang_test exp/tdnn exp/tdnn/graph\n",
+            "utils/mkgraph.sh --self-loop-scale 1.0 data/lang_test_adapt exp/tdnn exp/tdnn/graph_adapt\n",
+            "\n",
+            "# Lookahead part goes OOM\n",
+            "#utils/mkgraph_lookahead.sh \\\n",
+            "#          --self-loop-scale 1.0 data/lang \\\n",
+            "#          exp/tdnn data/en-us-mix.lm.gz exp/tdnn/lgraph\n",
+            "#utils/mkgraph_lookahead.sh \\\n",
+            "#          --self-loop-scale 1.0 data/lang \\\n",
+            "#          exp/tdnn db/en-50k-0.4-android.lm.gz exp/tdnn/lgraph_orig\n",
+            "+ . path.sh\n",
+            "+++ pwd\n",
+            "++ export KALDI_ROOT=/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../..\n",
+            "++ KALDI_ROOT=/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../..\n",
+            "++ export PATH=/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/utils:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../tools/openfst/bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/fstbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/gmmbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/featbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/lm:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/sgmmbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/sgmm2bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/fgmmbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/latbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/nnetbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/nnet2bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/online2bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/ivectorbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/lmbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/chainbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/nnet3bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab:/opt/bin:/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/tools/node/bin:/tools/google-cloud-sdk/bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../tools/sph2pipe_v2.5\n",
+            "++ PATH=/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/utils:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../tools/openfst/bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/fstbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/gmmbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/featbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/lm:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/sgmmbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/sgmm2bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/fgmmbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/latbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/nnetbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/nnet2bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/online2bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/ivectorbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/lmbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/chainbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/nnet3bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab:/opt/bin:/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/tools/node/bin:/tools/google-cloud-sdk/bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../tools/sph2pipe_v2.5\n",
+            "++ export PATH=/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../tools/ngram-1.3.7/src/bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/utils:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../tools/openfst/bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/fstbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/gmmbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/featbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/lm:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/sgmmbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/sgmm2bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/fgmmbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/latbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/nnetbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/nnet2bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/online2bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/ivectorbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/lmbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/chainbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/nnet3bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab:/opt/bin:/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/tools/node/bin:/tools/google-cloud-sdk/bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../tools/sph2pipe_v2.5\n",
+            "++ PATH=/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../tools/ngram-1.3.7/src/bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/utils:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../tools/openfst/bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/fstbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/gmmbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/featbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/lm:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/sgmmbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/sgmm2bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/fgmmbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/latbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/nnetbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/nnet2bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/online2bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/ivectorbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/lmbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/chainbin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../src/nnet3bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab:/opt/bin:/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/tools/node/bin:/tools/google-cloud-sdk/bin:/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../tools/sph2pipe_v2.5\n",
+            "++ export LD_LIBRARY_PATH=/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../tools/openfst/lib/fst/\n",
+            "++ LD_LIBRARY_PATH=/content/kaldi/egs/ac/vosk-model-small-en-us-0.15-compile-colab/../../../tools/openfst/lib/fst/\n",
+            "++ export LC_ALL=C\n",
+            "++ LC_ALL=C\n",
+            "+ pip3 install phonetisaurus\n",
+            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
+            "Requirement already satisfied: phonetisaurus in /usr/local/lib/python3.7/dist-packages (0.3.0)\n",
+            "+ rm -rf data\n",
+            "+ rm -rf exp/tdnn/lgraph\n",
+            "+ rm -rf exp/tdnn/lgraph_orig\n",
+            "+ mkdir -p data/dict\n",
+            "+ cp db/phone/extra_questions.txt db/phone/nonsilence_phones.txt db/phone/optional_silence.txt db/phone/silence_phones.txt data/dict\n",
+            "+ ./dict.py\n",
+            "+ python3 ./get_vocab.py\n",
+            "+ ngramsymbols data/mix.vocab data/mix.syms\n",
+            "+ farcompilestrings --fst_type=compact --symbols=data/mix.syms --keep_symbols '--unknown_symbol=[unk]' db/extra.txt data/extra.far\n",
+            "+ ngramcount --order=3 data/extra.far -\n",
+            "+ ngrammake --method=witten_bell - data/extra.mod\n",
+            "+ ngramshrink --method=count_prune --count_pattern=3+:3\n",
+            "+ ngramread\n",
+            "+ ngramprint --integers\n",
+            "+ grep -v '<unk>'\n",
+            "+ ngramread --renormalize_arpa --ARPA --symbols=data/mix.syms - data/en-us.mod\n",
+            "+ gunzip -c db/en-50k-0.4-android.lm.gz\n",
+            "+ ngrammerge --method=bayes_model_merge --normalize --alpha=0.95 --beta=0.05 data/en-us.mod data/extra.mod data/en-us-mix.mod\n",
+            "+ ngramprint --ARPA data/en-us-mix.mod\n",
+            "+ gzip -c\n",
+            "+ utils/prepare_lang.sh data/dict '[unk]' data/lang_local data/lang\n",
+            "utils/prepare_lang.sh data/dict [unk] data/lang_local data/lang\n",
+            "Checking data/dict/silence_phones.txt ...\n",
+            "--> reading data/dict/silence_phones.txt\n",
+            "--> text seems to be UTF-8 or ASCII, checking whitespaces\n",
+            "--> text contains only allowed whitespaces\n",
+            "--> data/dict/silence_phones.txt is OK\n",
+            "\n",
+            "Checking data/dict/optional_silence.txt ...\n",
+            "--> reading data/dict/optional_silence.txt\n",
+            "--> text seems to be UTF-8 or ASCII, checking whitespaces\n",
+            "--> text contains only allowed whitespaces\n",
+            "--> data/dict/optional_silence.txt is OK\n",
+            "\n",
+            "Checking data/dict/nonsilence_phones.txt ...\n",
+            "--> reading data/dict/nonsilence_phones.txt\n",
+            "--> text seems to be UTF-8 or ASCII, checking whitespaces\n",
+            "--> text contains only allowed whitespaces\n",
+            "--> data/dict/nonsilence_phones.txt is OK\n",
+            "\n",
+            "Checking disjoint: silence_phones.txt, nonsilence_phones.txt\n",
+            "--> disjoint property is OK.\n",
+            "\n",
+            "Checking data/dict/lexicon.txt\n",
+            "--> reading data/dict/lexicon.txt\n",
+            "--> text seems to be UTF-8 or ASCII, checking whitespaces\n",
+            "--> text contains only allowed whitespaces\n",
+            "--> data/dict/lexicon.txt is OK\n",
+            "\n",
+            "Checking data/dict/extra_questions.txt ...\n",
+            "--> data/dict/extra_questions.txt is empty (this is OK)\n",
+            "--> SUCCESS [validating dictionary directory data/dict]\n",
+            "\n",
+            "**Creating data/dict/lexiconp.txt from data/dict/lexicon.txt\n",
+            "fstaddselfloops data/lang/phones/wdisambig_phones.int data/lang/phones/wdisambig_words.int \n",
+            "prepare_lang.sh: validating output directory\n",
+            "utils/validate_lang.pl data/lang\n",
+            "Checking existence of separator file\n",
+            "separator file data/lang/subword_separator.txt is empty or does not exist, deal in word case.\n",
+            "Checking data/lang/phones.txt ...\n",
+            "--> text seems to be UTF-8 or ASCII, checking whitespaces\n",
+            "--> text contains only allowed whitespaces\n",
+            "--> data/lang/phones.txt is OK\n",
+            "\n",
+            "Checking words.txt: #0 ...\n",
+            "--> text seems to be UTF-8 or ASCII, checking whitespaces\n",
+            "--> text contains only allowed whitespaces\n",
+            "--> data/lang/words.txt is OK\n",
+            "\n",
+            "Checking disjoint: silence.txt, nonsilence.txt, disambig.txt ...\n",
+            "--> silence.txt and nonsilence.txt are disjoint\n",
+            "--> silence.txt and disambig.txt are disjoint\n",
+            "--> disambig.txt and nonsilence.txt are disjoint\n",
+            "--> disjoint property is OK\n",
+            "\n",
+            "Checking sumation: silence.txt, nonsilence.txt, disambig.txt ...\n",
+            "--> found no unexplainable phones in phones.txt\n",
+            "\n",
+            "Checking data/lang/phones/context_indep.{txt, int, csl} ...\n",
+            "--> text seems to be UTF-8 or ASCII, checking whitespaces\n",
+            "--> text contains only allowed whitespaces\n",
+            "--> 10 entry/entries in data/lang/phones/context_indep.txt\n",
+            "--> data/lang/phones/context_indep.int corresponds to data/lang/phones/context_indep.txt\n",
+            "--> data/lang/phones/context_indep.csl corresponds to data/lang/phones/context_indep.txt\n",
+            "--> data/lang/phones/context_indep.{txt, int, csl} are OK\n",
+            "\n",
+            "Checking data/lang/phones/nonsilence.{txt, int, csl} ...\n",
+            "--> text seems to be UTF-8 or ASCII, checking whitespaces\n",
+            "--> text contains only allowed whitespaces\n",
+            "--> 156 entry/entries in data/lang/phones/nonsilence.txt\n",
+            "--> data/lang/phones/nonsilence.int corresponds to data/lang/phones/nonsilence.txt\n",
+            "--> data/lang/phones/nonsilence.csl corresponds to data/lang/phones/nonsilence.txt\n",
+            "--> data/lang/phones/nonsilence.{txt, int, csl} are OK\n",
+            "\n",
+            "Checking data/lang/phones/silence.{txt, int, csl} ...\n",
+            "--> text seems to be UTF-8 or ASCII, checking whitespaces\n",
+            "--> text contains only allowed whitespaces\n",
+            "--> 10 entry/entries in data/lang/phones/silence.txt\n",
+            "--> data/lang/phones/silence.int corresponds to data/lang/phones/silence.txt\n",
+            "--> data/lang/phones/silence.csl corresponds to data/lang/phones/silence.txt\n",
+            "--> data/lang/phones/silence.{txt, int, csl} are OK\n",
+            "\n",
+            "Checking data/lang/phones/optional_silence.{txt, int, csl} ...\n",
+            "--> text seems to be UTF-8 or ASCII, checking whitespaces\n",
+            "--> text contains only allowed whitespaces\n",
+            "--> 1 entry/entries in data/lang/phones/optional_silence.txt\n",
+            "--> data/lang/phones/optional_silence.int corresponds to data/lang/phones/optional_silence.txt\n",
+            "--> data/lang/phones/optional_silence.csl corresponds to data/lang/phones/optional_silence.txt\n",
+            "--> data/lang/phones/optional_silence.{txt, int, csl} are OK\n",
+            "\n",
+            "Checking data/lang/phones/disambig.{txt, int, csl} ...\n",
+            "--> text seems to be UTF-8 or ASCII, checking whitespaces\n",
+            "--> text contains only allowed whitespaces\n",
+            "--> 14 entry/entries in data/lang/phones/disambig.txt\n",
+            "--> data/lang/phones/disambig.int corresponds to data/lang/phones/disambig.txt\n",
+            "--> data/lang/phones/disambig.csl corresponds to data/lang/phones/disambig.txt\n",
+            "--> data/lang/phones/disambig.{txt, int, csl} are OK\n",
+            "\n",
+            "Checking data/lang/phones/roots.{txt, int} ...\n",
+            "--> text seems to be UTF-8 or ASCII, checking whitespaces\n",
+            "--> text contains only allowed whitespaces\n",
+            "--> 41 entry/entries in data/lang/phones/roots.txt\n",
+            "--> data/lang/phones/roots.int corresponds to data/lang/phones/roots.txt\n",
+            "--> data/lang/phones/roots.{txt, int} are OK\n",
+            "\n",
+            "Checking data/lang/phones/sets.{txt, int} ...\n",
+            "--> text seems to be UTF-8 or ASCII, checking whitespaces\n",
+            "--> text contains only allowed whitespaces\n",
+            "--> 41 entry/entries in data/lang/phones/sets.txt\n",
+            "--> data/lang/phones/sets.int corresponds to data/lang/phones/sets.txt\n",
+            "--> data/lang/phones/sets.{txt, int} are OK\n",
+            "\n",
+            "Checking data/lang/phones/extra_questions.{txt, int} ...\n",
+            "--> text seems to be UTF-8 or ASCII, checking whitespaces\n",
+            "--> text contains only allowed whitespaces\n",
+            "--> 9 entry/entries in data/lang/phones/extra_questions.txt\n",
+            "--> data/lang/phones/extra_questions.int corresponds to data/lang/phones/extra_questions.txt\n",
+            "--> data/lang/phones/extra_questions.{txt, int} are OK\n",
+            "\n",
+            "Checking data/lang/phones/word_boundary.{txt, int} ...\n",
+            "--> text seems to be UTF-8 or ASCII, checking whitespaces\n",
+            "--> text contains only allowed whitespaces\n",
+            "--> 166 entry/entries in data/lang/phones/word_boundary.txt\n",
+            "--> data/lang/phones/word_boundary.int corresponds to data/lang/phones/word_boundary.txt\n",
+            "--> data/lang/phones/word_boundary.{txt, int} are OK\n",
+            "\n",
+            "Checking optional_silence.txt ...\n",
+            "--> reading data/lang/phones/optional_silence.txt\n",
+            "--> data/lang/phones/optional_silence.txt is OK\n",
+            "\n",
+            "Checking disambiguation symbols: #0 and #1\n",
+            "--> data/lang/phones/disambig.txt has \"#0\" and \"#1\"\n",
+            "--> data/lang/phones/disambig.txt is OK\n",
+            "\n",
+            "Checking topo ...\n",
+            "\n",
+            "Checking word_boundary.txt: silence.txt, nonsilence.txt, disambig.txt ...\n",
+            "--> data/lang/phones/word_boundary.txt doesn't include disambiguation symbols\n",
+            "--> data/lang/phones/word_boundary.txt is the union of nonsilence.txt and silence.txt\n",
+            "--> data/lang/phones/word_boundary.txt is OK\n",
+            "\n",
+            "Checking word-level disambiguation symbols...\n",
+            "--> data/lang/phones/wdisambig.txt exists (newer prepare_lang.sh)\n",
+            "Checking word_boundary.int and disambig.int\n",
+            "--> generating a 98 word/subword sequence\n",
+            "--> resulting phone sequence from L.fst corresponds to the word sequence\n",
+            "--> L.fst is OK\n",
+            "--> generating a 49 word/subword sequence\n",
+            "--> resulting phone sequence from L_disambig.fst corresponds to the word sequence\n",
+            "--> L_disambig.fst is OK\n",
+            "\n",
+            "Checking data/lang/oov.{txt, int} ...\n",
+            "--> text seems to be UTF-8 or ASCII, checking whitespaces\n",
+            "--> text contains only allowed whitespaces\n",
+            "--> 1 entry/entries in data/lang/oov.txt\n",
+            "--> data/lang/oov.int corresponds to data/lang/oov.txt\n",
+            "--> data/lang/oov.{txt, int} are OK\n",
+            "\n",
+            "--> data/lang/L.fst is olabel sorted\n",
+            "--> data/lang/L_disambig.fst is olabel sorted\n",
+            "--> SUCCESS [validating lang directory data/lang]\n",
+            "+ utils/format_lm.sh data/lang db/en-50k-0.4-android.lm.gz data/dict/lexicon.txt data/lang_test\n",
+            "Converting 'db/en-50k-0.4-android.lm.gz' to FST\n",
+            "arpa2fst --disambig-symbol=#0 --read-symbol-table=data/lang_test/words.txt - data/lang_test/G.fst \n",
+            "LOG (arpa2fst[5.5.1046~1-76cd5]:Read():arpa-file-parser.cc:94) Reading \\data\\ section.\n",
+            "LOG (arpa2fst[5.5.1046~1-76cd5]:Read():arpa-file-parser.cc:149) Reading \\1-grams: section.\n",
+            "LOG (arpa2fst[5.5.1046~1-76cd5]:Read():arpa-file-parser.cc:149) Reading \\2-grams: section.\n",
+            "LOG (arpa2fst[5.5.1046~1-76cd5]:Read():arpa-file-parser.cc:149) Reading \\3-grams: section.\n",
+            "LOG (arpa2fst[5.5.1046~1-76cd5]:RemoveRedundantStates():arpa-lm-compiler.cc:359) Reduced num-states from 1217362 to 185036\n",
+            "fstisstochastic data/lang_test/G.fst \n",
+            "0.476411 -3.03779\n",
+            "Succeeded in formatting LM: 'db/en-50k-0.4-android.lm.gz'\n",
+            "+ utils/format_lm.sh data/lang data/en-us-mix.lm.gz data/dict/lexicon.txt data/lang_test_adapt\n",
+            "Converting 'data/en-us-mix.lm.gz' to FST\n",
+            "arpa2fst --disambig-symbol=#0 --read-symbol-table=data/lang_test_adapt/words.txt - data/lang_test_adapt/G.fst \n",
+            "LOG (arpa2fst[5.5.1046~1-76cd5]:Read():arpa-file-parser.cc:94) Reading \\data\\ section.\n",
+            "LOG (arpa2fst[5.5.1046~1-76cd5]:Read():arpa-file-parser.cc:149) Reading \\1-grams: section.\n",
+            "LOG (arpa2fst[5.5.1046~1-76cd5]:Read():arpa-file-parser.cc:149) Reading \\2-grams: section.\n",
+            "LOG (arpa2fst[5.5.1046~1-76cd5]:Read():arpa-file-parser.cc:149) Reading \\3-grams: section.\n",
+            "LOG (arpa2fst[5.5.1046~1-76cd5]:RemoveRedundantStates():arpa-lm-compiler.cc:359) Reduced num-states from 1217646 to 185095\n",
+            "fstisstochastic data/lang_test_adapt/G.fst \n",
+            "6.81902e-07 -3.03779\n",
+            "Succeeded in formatting LM: 'data/en-us-mix.lm.gz'\n",
+            "+ utils/mkgraph.sh --self-loop-scale 1.0 data/lang_test exp/tdnn exp/tdnn/graph\n",
+            "tree-info exp/tdnn/tree \n",
+            "tree-info exp/tdnn/tree \n",
+            "fstdeterminizestar --use-log=true \n",
+            "fsttablecompose data/lang_test/L_disambig.fst data/lang_test/G.fst \n",
+            "fstminimizeencoded \n",
+            "fstpushspecial \n",
+            "fstisstochastic data/lang_test/tmp/LG.fst \n",
+            "-0.145498 -0.146281\n",
+            "[info]: LG not stochastic.\n",
+            "fstcomposecontext --context-size=2 --central-position=1 --read-disambig-syms=data/lang_test/phones/disambig.int --write-disambig-syms=data/lang_test/tmp/disambig_ilabels_2_1.int data/lang_test/tmp/ilabels_2_1.905 data/lang_test/tmp/LG.fst \n",
+            "fstisstochastic data/lang_test/tmp/CLG_2_1.fst \n",
+            "-0.145498 -0.146281\n",
+            "[info]: CLG not stochastic.\n",
+            "make-h-transducer --disambig-syms-out=exp/tdnn/graph/disambig_tid.int --transition-scale=1.0 data/lang_test/tmp/ilabels_2_1 exp/tdnn/tree exp/tdnn/final.mdl \n",
+            "fstrmepslocal \n",
+            "fsttablecompose exp/tdnn/graph/Ha.fst data/lang_test/tmp/CLG_2_1.fst \n",
+            "fstdeterminizestar --use-log=true \n",
+            "fstminimizeencoded \n",
+            "fstrmsymbols exp/tdnn/graph/disambig_tid.int \n",
+            "fstisstochastic exp/tdnn/graph/HCLGa.fst \n",
+            "-0.109817 -0.571742\n",
+            "HCLGa is not stochastic\n",
+            "add-self-loops --self-loop-scale=1.0 --reorder=true exp/tdnn/final.mdl exp/tdnn/graph/HCLGa.fst \n",
+            "fstisstochastic exp/tdnn/graph/HCLG.fst \n",
+            "1.90465e-09 -0.415046\n",
+            "[info]: final HCLG is not stochastic.\n",
+            "+ utils/mkgraph.sh --self-loop-scale 1.0 data/lang_test_adapt exp/tdnn exp/tdnn/graph_adapt\n",
+            "tree-info exp/tdnn/tree \n",
+            "tree-info exp/tdnn/tree \n",
+            "fstdeterminizestar --use-log=true \n",
+            "fsttablecompose data/lang_test_adapt/L_disambig.fst data/lang_test_adapt/G.fst \n",
+            "fstminimizeencoded \n",
+            "fstpushspecial \n",
+            "fstisstochastic data/lang_test_adapt/tmp/LG.fst \n",
+            "-0.148474 -0.149181\n",
+            "[info]: LG not stochastic.\n",
+            "fstcomposecontext --context-size=2 --central-position=1 --read-disambig-syms=data/lang_test_adapt/phones/disambig.int --write-disambig-syms=data/lang_test_adapt/tmp/disambig_ilabels_2_1.int data/lang_test_adapt/tmp/ilabels_2_1.979 data/lang_test_adapt/tmp/LG.fst \n",
+            "fstisstochastic data/lang_test_adapt/tmp/CLG_2_1.fst \n",
+            "-0.148474 -0.149181\n",
+            "[info]: CLG not stochastic.\n",
+            "make-h-transducer --disambig-syms-out=exp/tdnn/graph_adapt/disambig_tid.int --transition-scale=1.0 data/lang_test_adapt/tmp/ilabels_2_1 exp/tdnn/tree exp/tdnn/final.mdl \n",
+            "fstrmepslocal \n",
+            "fsttablecompose exp/tdnn/graph_adapt/Ha.fst data/lang_test_adapt/tmp/CLG_2_1.fst \n",
+            "fstdeterminizestar --use-log=true \n",
+            "fstminimizeencoded \n",
+            "fstrmsymbols exp/tdnn/graph_adapt/disambig_tid.int \n",
+            "fstisstochastic exp/tdnn/graph_adapt/HCLGa.fst \n",
+            "-0.113907 -0.5857\n",
+            "HCLGa is not stochastic\n",
+            "add-self-loops --self-loop-scale=1.0 --reorder=true exp/tdnn/final.mdl exp/tdnn/graph_adapt/HCLGa.fst \n",
+            "fstisstochastic exp/tdnn/graph_adapt/HCLG.fst \n",
+            "1.90465e-09 -0.423618\n",
+            "[info]: final HCLG is not stochastic.\n"
+          ]
+        }
+      ]
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "!cat decode.sh\n",
+        "!bash decode.sh"
+      ],
+      "metadata": {
+        "colab": {
+          "base_uri": "https://localhost:8080/"
+        },
+        "id": "Sl3QBI1MXpc-",
+        "outputId": "affac8a3-782f-4000-e31f-81bfed47a37a"
+      },
+      "execution_count": 10,
+      "outputs": [
+        {
+          "output_type": "stream",
+          "name": "stdout",
+          "text": [
+            "#!/bin/bash\n",
+            "\n",
+            ". path.sh\n",
+            "\n",
+            "steps/make_mfcc.sh --nj 10 data_test/test_small exp/make_mfcc/test mfcc\n",
+            "steps/compute_cmvn_stats.sh data_test/test_small exp/make_mfcc/test mfcc\n",
+            "utils/fix_data_dir.sh data_test/test_small\n",
+            "\n",
+            "steps/online/nnet2/extract_ivectors_online.sh --nj 4 \\\n",
+            "          data_test/test_small exp/extractor \\\n",
+            "          exp/ivectors_test\n",
+            "\n",
+            "steps/nnet3/decode.sh --nj 4 \\\n",
+            "          --acwt 1.0 --post-decode-acwt 10.0 \\\n",
+            "          --online-ivector-dir exp/ivectors_test \\\n",
+            "          exp/tdnn/graph_adapt data_test/test_small exp/tdnn/decode_test_adapt\n",
+            "\n",
+            "steps/nnet3/decode.sh --nj 4 \\\n",
+            "          --acwt 1.0 --post-decode-acwt 10.0 \\\n",
+            "          --online-ivector-dir exp/ivectors_test \\\n",
+            "          exp/tdnn/graph data_test/test_small exp/tdnn/decode_test\n",
+            "\n",
+            "#steps/nnet3/decode_lookahead.sh --nj 4 \\\n",
+            "#          --acwt 1.0 --post-decode-acwt 10.0 \\\n",
+            "#          --online-ivector-dir exp/ivectors_test \\\n",
+            "#          exp/tdnn/lgraph data_test/test_small exp/tdnn/decode_test_adapt\n",
+            "#steps/nnet3/decode_lookahead.sh --nj 4 \\\n",
+            "#          --acwt 1.0 --post-decode-acwt 10.0 \\\n",
+            "#          --online-ivector-dir exp/ivectors_test \\\n",
+            "#          exp/tdnn/lgraph_orig data_test/test_small exp/tdnn/decode_test\n",
+            "steps/make_mfcc.sh --nj 10 data_test/test_small exp/make_mfcc/test mfcc\n",
+            "steps/make_mfcc.sh: moving data_test/test_small/feats.scp to data_test/test_small/.backup\n",
+            "utils/validate_data_dir.sh: Successfully validated data-directory data_test/test_small\n",
+            "steps/make_mfcc.sh: [info]: no segments file exists: assuming wav.scp indexed by utterance.\n",
+            "steps/make_mfcc.sh: Succeeded creating MFCC features for test_small\n",
+            "steps/compute_cmvn_stats.sh data_test/test_small exp/make_mfcc/test mfcc\n",
+            "Succeeded creating CMVN stats for test_small\n",
+            "fix_data_dir.sh: kept all 50 utterances.\n",
+            "fix_data_dir.sh: old files are kept in data_test/test_small/.backup\n",
+            "steps/online/nnet2/extract_ivectors_online.sh --nj 4 data_test/test_small exp/extractor exp/ivectors_test\n",
+            "steps/online/nnet2/extract_ivectors_online.sh: extracting iVectors\n",
+            "steps/online/nnet2/extract_ivectors_online.sh: combining iVectors across jobs\n",
+            "steps/online/nnet2/extract_ivectors_online.sh: done extracting (online) iVectors to exp/ivectors_test using the extractor in exp/extractor.\n",
+            "steps/nnet3/decode.sh --nj 4 --acwt 1.0 --post-decode-acwt 10.0 --online-ivector-dir exp/ivectors_test exp/tdnn/graph_adapt data_test/test_small exp/tdnn/decode_test_adapt\n",
+            "steps/nnet2/check_ivectors_compatible.sh: WARNING: One of the directories do not contain iVector ID.\n",
+            "steps/nnet2/check_ivectors_compatible.sh: WARNING: That means it's you who's reponsible for keeping \n",
+            "steps/nnet2/check_ivectors_compatible.sh: WARNING: the directories compatible\n",
+            "steps/nnet3/decode.sh: feature type is raw\n",
+            "steps/diagnostic/analyze_lats.sh --cmd run.pl --iter final exp/tdnn/graph_adapt exp/tdnn/decode_test_adapt\n",
+            "steps/diagnostic/analyze_lats.sh: see stats in exp/tdnn/decode_test_adapt/log/analyze_alignments.log\n",
+            "Overall, lattice depth (10,50,90-percentile)=(1,1,4) and mean=2.4\n",
+            "steps/diagnostic/analyze_lats.sh: see stats in exp/tdnn/decode_test_adapt/log/analyze_lattice_depth_stats.log\n",
+            "score best paths\n",
+            "local/score.sh --cmd run.pl data_test/test_small exp/tdnn/graph_adapt exp/tdnn/decode_test_adapt\n",
+            "local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0\n",
+            "score confidence and timing with sclite\n",
+            "Decoding done.\n",
+            "steps/nnet3/decode.sh --nj 4 --acwt 1.0 --post-decode-acwt 10.0 --online-ivector-dir exp/ivectors_test exp/tdnn/graph data_test/test_small exp/tdnn/decode_test\n",
+            "steps/nnet2/check_ivectors_compatible.sh: WARNING: One of the directories do not contain iVector ID.\n",
+            "steps/nnet2/check_ivectors_compatible.sh: WARNING: That means it's you who's reponsible for keeping \n",
+            "steps/nnet2/check_ivectors_compatible.sh: WARNING: the directories compatible\n",
+            "steps/nnet3/decode.sh: feature type is raw\n",
+            "steps/diagnostic/analyze_lats.sh --cmd run.pl --iter final exp/tdnn/graph exp/tdnn/decode_test\n",
+            "steps/diagnostic/analyze_lats.sh: see stats in exp/tdnn/decode_test/log/analyze_alignments.log\n",
+            "Overall, lattice depth (10,50,90-percentile)=(1,5,23) and mean=10.4\n",
+            "steps/diagnostic/analyze_lats.sh: see stats in exp/tdnn/decode_test/log/analyze_lattice_depth_stats.log\n",
+            "score best paths\n",
+            "local/score.sh --cmd run.pl data_test/test_small exp/tdnn/graph exp/tdnn/decode_test\n",
+            "local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0\n",
+            "score confidence and timing with sclite\n",
+            "Decoding done.\n"
+          ]
+        }
+      ]
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "!bash RESULTS"
+      ],
+      "metadata": {
+        "id": "ABtcNyUDX4S8",
+        "outputId": "d5e50be7-3293-4a59-94b8-9bfa46736481",
+        "colab": {
+          "base_uri": "https://localhost:8080/"
+        }
+      },
+      "execution_count": 11,
+      "outputs": [
+        {
+          "output_type": "stream",
+          "name": "stdout",
+          "text": [
+            "%WER 11.77 [ 107 / 909, 13 ins, 7 del, 87 sub ] exp/tdnn/decode_test/wer_7_1.0\n",
+            "%WER 0.22 [ 2 / 909, 0 ins, 1 del, 1 sub ] exp/tdnn/decode_test_adapt/wer_10_1.0\n"
+          ]
+        }
+      ]
+    }
+  ]
+}

python/example/test_alternatives.py

@@ -1,17 +1,17 @@
 #!/usr/bin/env python3
 
-from vosk import Model, KaldiRecognizer, SetLogLevel
-import sys
-import os
 import wave
+import sys
 import json
 
+from vosk import Model, KaldiRecognizer, SetLogLevel
+
 SetLogLevel(0)
 
 wf = wave.open(sys.argv[1], "rb")
 if wf.getnchannels() != 1 or wf.getsampwidth() != 2 or wf.getcomptype() != "NONE":
-    print ("Audio file must be WAV format mono PCM.")
-    exit (1)
+    print("Audio file must be WAV format mono PCM.")
+    sys.exit(1)
 
 model = Model(lang="en-us")
 rec = KaldiRecognizer(model, wf.getframerate())

python/example/test_empty.py

@@ -1,11 +1,11 @@
 #!/usr/bin/env python3
 
-from vosk import Model, KaldiRecognizer
-import sys
 import json
 
+from vosk import Model, KaldiRecognizer
+
 model = Model(lang="en-us")
 rec = KaldiRecognizer(model, 8000)
 
 res = json.loads(rec.FinalResult())
-print (res)
+print(res)

python/example/test_ffmpeg.py

@@ -1,29 +1,29 @@
 #!/usr/bin/env python3
 
-from vosk import Model, KaldiRecognizer, SetLogLevel
-import sys
-import os
-import wave
 import subprocess
+import sys
+
+from vosk import Model, KaldiRecognizer, SetLogLevel
+
+SAMPLE_RATE = 16000
 
 SetLogLevel(0)
 
-sample_rate=16000
 model = Model(lang="en-us")
-rec = KaldiRecognizer(model, sample_rate)
+rec = KaldiRecognizer(model, SAMPLE_RATE)
 
-process = subprocess.Popen(['ffmpeg', '-loglevel', 'quiet', '-i',
+with subprocess.Popen(["ffmpeg", "-loglevel", "quiet", "-i",
                             sys.argv[1],
-                            '-ar', str(sample_rate) , '-ac', '1', '-f', 's16le', '-'],
-                            stdout=subprocess.PIPE)
+                            "-ar", str(SAMPLE_RATE) , "-ac", "1", "-f", "s16le", "-"],
+                            stdout=subprocess.PIPE) as process:
 
-while True:
-    data = process.stdout.read(4000)
-    if len(data) == 0:
-        break
-    if rec.AcceptWaveform(data):
-        print(rec.Result())
-    else:
-        print(rec.PartialResult())
+    while True:
+        data = process.stdout.read(4000)
+        if len(data) == 0:
+            break
+        if rec.AcceptWaveform(data):
+            print(rec.Result())
+        else:
+            print(rec.PartialResult())
 
-print(rec.FinalResult())
+    print(rec.FinalResult())

python/example/test_gpu_batch.py

@@ -1,26 +1,25 @@
 #!/usr/bin/env python3
 
 import sys
-import os
-import wave
-from time import sleep
 import json
-from timeit import default_timer as timer
-
 
 from vosk import BatchModel, BatchRecognizer, GpuInit
+from timeit import default_timer as timer
+
+TOT_SAMPLES = 0
 
 GpuInit()
 
-model = BatchModel()
+model = BatchModel("model")
+
+with open(sys.argv[1]) as fn:
+    fnames = fn.readlines()
+    fds = [open(x.strip(), "rb") for x in fnames]
+    uids = [fname.strip().split("/")[-1][:-4] for fname in fnames]
+    recs = [BatchRecognizer(model, 16000) for x in fnames]
+    results = [""] * len(fnames)
 
-fnames = open(sys.argv[1]).readlines()
-fds = [open(x.strip(), "rb") for x in fnames]
-uids = [fname.strip().split('/')[-1][:-4] for fname in fnames]
-recs = [BatchRecognizer(model, 16000) for x in fnames]
-results = [""] * len(fnames)
 ended = set()
-tot_samples = 0
 
 start_time = timer()
 
@@ -36,24 +35,27 @@ while True:
             ended.add(i)
             continue
         recs[i].AcceptWaveform(data)
-        tot_samples += len(data)
+        TOT_SAMPLES += len(data)
 
     # Wait for results from CUDA
     model.Wait()
 
     # Retrieve and add results
     for i, fd in enumerate(fds):
-       res = recs[i].Result()
-       if len(res) != 0:
-           results[i] = results[i] + " " + json.loads(res)['text']
+        res = recs[i].Result()
+        if len(res) != 0:
+            results[i] = results[i] + " " + json.loads(res)["text"]
 
     if len(ended) == len(fds):
         break
 
 end_time = timer()
 
-for i in range(len(results)):
-    print (uids[i], results[i].strip())
+for i, res in enumerate(results):
+    print(uids[i], res.strip())
 
-print ("Processed %.3f seconds of audio in %.3f seconds (%.3f xRT)" % (tot_samples / 16000.0 / 2, end_time - start_time, 
-    (tot_samples / 16000.0 / 2 / (end_time - start_time))), file=sys.stderr)
+print("Processed %.3f seconds of audio in %.3f seconds (%.3f xRT)"
+    % (TOT_SAMPLES / 16000.0 / 2,
+    end_time - start_time,
+    (TOT_SAMPLES / 16000.0 / 2 / (end_time - start_time))),
+    file=sys.stderr)

python/example/test_gradio.py

@@ -0,0 +1,40 @@
+#!/usr/bin/env python3
+
+import json
+import gradio as gr
+
+from vosk import KaldiRecognizer, Model
+
+model = Model(lang="en-us")
+
+def transcribe(data, state):
+    sample_rate, audio_data = data
+    audio_data = (audio_data >> 16).astype("int16").tobytes()
+
+    if state is None:
+        rec = KaldiRecognizer(model, sample_rate)
+        result = []
+    else:
+        rec, result = state
+
+    if rec.AcceptWaveform(audio_data):
+        text_result = json.loads(rec.Result())["text"]
+        if text_result != "":
+            result.append(text_result)
+        partial_result = ""
+    else:
+        partial_result = json.loads(rec.PartialResult())["partial"] + " "
+
+    return "\n".join(result) + "\n" + partial_result, (rec, result)
+
+gr.Interface(
+    fn=transcribe,
+    inputs=[
+        gr.Audio(source="microphone", type="numpy", streaming=True),
+        "state"
+    ],
+    outputs=[
+        "textbox",
+        "state"
+    ],
+    live=True).launch(share=True)

python/example/test_microphone.py

@@ -1,11 +1,15 @@
 #!/usr/bin/env python3
 
+# prerequisites: as described in https://alphacephei.com/vosk/install and also python module `sounddevice` (simply run command `pip install sounddevice`)
+# Example usage using Dutch (nl) recognition model: `python test_microphone.py -m nl`
+# For more help run: `python test_microphone.py -h`
+
 import argparse
-import os
 import queue
-import sounddevice as sd
-import vosk
 import sys
+import sounddevice as sd
+
+from vosk import Model, KaldiRecognizer
 
 q = queue.Queue()
 
@@ -24,8 +28,8 @@ def callback(indata, frames, time, status):
 
 parser = argparse.ArgumentParser(add_help=False)
 parser.add_argument(
-    '-l', '--list-devices', action='store_true',
-    help='show list of audio devices and exit')
+    "-l", "--list-devices", action="store_true",
+    help="show list of audio devices and exit")
 args, remaining = parser.parse_known_args()
 if args.list_devices:
     print(sd.query_devices())
@@ -35,46 +39,51 @@ parser = argparse.ArgumentParser(
     formatter_class=argparse.RawDescriptionHelpFormatter,
     parents=[parser])
 parser.add_argument(
-    '-f', '--filename', type=str, metavar='FILENAME',
-    help='audio file to store recording to')
+    "-f", "--filename", type=str, metavar="FILENAME",
+    help="audio file to store recording to")
 parser.add_argument(
-    '-d', '--device', type=int_or_str,
-    help='input device (numeric ID or substring)')
+    "-d", "--device", type=int_or_str,
+    help="input device (numeric ID or substring)")
 parser.add_argument(
-    '-r', '--samplerate', type=int, help='sampling rate')
+    "-r", "--samplerate", type=int, help="sampling rate")
+parser.add_argument(
+    "-m", "--model", type=str, help="language model; e.g. en-us, fr, nl; default is en-us")
 args = parser.parse_args(remaining)
 
 try:
     if args.samplerate is None:
-        device_info = sd.query_devices(args.device, 'input')
+        device_info = sd.query_devices(args.device, "input")
         # soundfile expects an int, sounddevice provides a float:
-        args.samplerate = int(device_info['default_samplerate'])
-
-    model = vosk.Model(lang="en-us")
+        args.samplerate = int(device_info["default_samplerate"])
+        
+    if args.model is None:
+        model = Model(lang="en-us")
+    else:
+        model = Model(lang=args.model)
 
     if args.filename:
         dump_fn = open(args.filename, "wb")
     else:
         dump_fn = None
 
-    with sd.RawInputStream(samplerate=args.samplerate, blocksize = 8000, device=args.device, dtype='int16',
-                            channels=1, callback=callback):
-            print('#' * 80)
-            print('Press Ctrl+C to stop the recording')
-            print('#' * 80)
+    with sd.RawInputStream(samplerate=args.samplerate, blocksize = 8000, device=args.device,
+            dtype="int16", channels=1, callback=callback):
+        print("#" * 80)
+        print("Press Ctrl+C to stop the recording")
+        print("#" * 80)
 
-            rec = vosk.KaldiRecognizer(model, args.samplerate)
-            while True:
-                data = q.get()
-                if rec.AcceptWaveform(data):
-                    print(rec.Result())
-                else:
-                    print(rec.PartialResult())
-                if dump_fn is not None:
-                    dump_fn.write(data)
+        rec = KaldiRecognizer(model, args.samplerate)
+        while True:
+            data = q.get()
+            if rec.AcceptWaveform(data):
+                print(rec.Result())
+            else:
+                print(rec.PartialResult())
+            if dump_fn is not None:
+                dump_fn.write(data)
 
 except KeyboardInterrupt:
-    print('\nDone')
+    print("\nDone")
     parser.exit(0)
 except Exception as e:
-    parser.exit(type(e).__name__ + ': ' + str(e))
+    parser.exit(type(e).__name__ + ": " + str(e))

python/example/test_nlsml.py

@@ -1,16 +1,16 @@
 #!/usr/bin/env python3
 
-from vosk import Model, KaldiRecognizer, SetLogLevel
-import sys
-import os
 import wave
+import sys
+
+from vosk import Model, KaldiRecognizer, SetLogLevel
 
 SetLogLevel(0)
 
 wf = wave.open(sys.argv[1], "rb")
 if wf.getnchannels() != 1 or wf.getsampwidth() != 2 or wf.getcomptype() != "NONE":
-    print ("Audio file must be WAV format mono PCM.")
-    exit (1)
+    print("Audio file must be WAV format mono PCM.")
+    sys.exit(1)
 
 model = Model(lang="en-us")
 rec = KaldiRecognizer(model, wf.getframerate())

python/example/test_reset.py

@@ -1,17 +1,17 @@
 #!/usr/bin/env python3
 
-from vosk import Model, KaldiRecognizer, SetLogLevel
-import sys
-import os
 import wave
+import sys
 import json
 
+from vosk import Model, KaldiRecognizer, SetLogLevel
+
 SetLogLevel(0)
 
 wf = wave.open(sys.argv[1], "rb")
 if wf.getnchannels() != 1 or wf.getsampwidth() != 2 or wf.getcomptype() != "NONE":
-    print ("Audio file must be WAV format mono PCM.")
-    exit (1)
+    print("Audio file must be WAV format mono PCM.")
+    sys.exit(1)
 
 model = Model(lang="en-us")
 rec = KaldiRecognizer(model, wf.getframerate())
@@ -22,12 +22,12 @@ while True:
         break
     if rec.AcceptWaveform(data):
         print(rec.Result())
-        break
+        sys.exit(1)
 
     else:
         jres = json.loads(rec.PartialResult())
         print(jres)
 
-        if jres['partial'] == "one zero zero zero":
+        if jres["partial"] == "one zero zero zero":
             print("We can reset recognizer here and start over")
-            rec.Reset();
+            rec.Reset()

python/example/test_simple.py

@@ -1,17 +1,17 @@
 #!/usr/bin/env python3
 
-from vosk import Model, KaldiRecognizer, SetLogLevel
-import sys
-import os
 import wave
+import sys
+
+from vosk import Model, KaldiRecognizer, SetLogLevel
 
 # You can set log level to -1 to disable debug messages
 SetLogLevel(0)
 
 wf = wave.open(sys.argv[1], "rb")
 if wf.getnchannels() != 1 or wf.getsampwidth() != 2 or wf.getcomptype() != "NONE":
-    print ("Audio file must be WAV format mono PCM.")
-    exit (1)
+    print("Audio file must be WAV format mono PCM.")
+    sys.exit(1)
 
 model = Model(lang="en-us")
 

python/example/test_speaker.py

@@ -1,33 +1,54 @@
 #!/usr/bin/env python3
 
-from vosk import Model, KaldiRecognizer, SpkModel
+import os
 import sys
 import wave
 import json
-import os
 import numpy as np
 
-spk_model_path = "model-spk"
+from vosk import Model, KaldiRecognizer, SpkModel
 
-if not os.path.exists(spk_model_path):
-    print ("Please download the speaker model from https://alphacephei.com/vosk/models and unpack as {} in the current folder.".format(spk_model_path))
-    exit (1)
+SPK_MODEL_PATH = "model-spk"
+
+if not os.path.exists(SPK_MODEL_PATH):
+    print("Please download the speaker model from "
+        "https://alphacephei.com/vosk/models and unpack as {SPK_MODEL_PATH} "
+        "in the current folder.")
+    sys.exit(1)
 
 wf = wave.open(sys.argv[1], "rb")
 if wf.getnchannels() != 1 or wf.getsampwidth() != 2 or wf.getcomptype() != "NONE":
-    print ("Audio file must be WAV format mono PCM.")
-    exit (1)
+    print("Audio file must be WAV format mono PCM.")
+    sys.exit(1)
 
 # Large vocabulary free form recognition
 model = Model(lang="en-us")
-spk_model = SpkModel(spk_model_path)
+spk_model = SpkModel(SPK_MODEL_PATH)
 #rec = KaldiRecognizer(model, wf.getframerate(), spk_model)
 rec = KaldiRecognizer(model, wf.getframerate())
 rec.SetSpkModel(spk_model)
 
-# We compare speakers with cosine distance. We can keep one or several fingerprints for the speaker in a database
+# We compare speakers with cosine distance.
+# We can keep one or several fingerprints for the speaker in a database
 # to distingusih among users.
-spk_sig = [-1.110417,0.09703002,1.35658,0.7798632,-0.305457,-0.339204,0.6186931,-0.4521213,0.3982236,-0.004530723,0.7651616,0.6500852,-0.6664245,0.1361499,0.1358056,-0.2887807,-0.1280468,-0.8208137,-1.620276,-0.4628615,0.7870904,-0.105754,0.9739769,-0.3258137,-0.7322628,-0.6212429,-0.5531687,-0.7796484,0.7035915,1.056094,-0.4941756,-0.6521456,-0.2238328,-0.003737517,0.2165709,1.200186,-0.7737719,0.492015,1.16058,0.6135428,-0.7183084,0.3153541,0.3458071,-1.418189,-0.9624157,0.4168292,-1.627305,0.2742135,-0.6166027,0.1962581,-0.6406527,0.4372789,-0.4296024,0.4898657,-0.9531326,-0.2945702,0.7879696,-1.517101,-0.9344181,-0.5049928,-0.005040941,-0.4637912,0.8223695,-1.079849,0.8871287,-0.9732434,-0.5548235,1.879138,-1.452064,-0.1975368,1.55047,0.5941782,-0.52897,1.368219,0.6782904,1.202505,-0.9256122,-0.9718158,-0.9570228,-0.5563112,-1.19049,-1.167985,2.606804,-2.261825,0.01340385,0.2526799,-1.125458,-1.575991,-0.363153,0.3270262,1.485984,-1.769565,1.541829,0.7293826,0.1743717,-0.4759418,1.523451,-2.487134,-1.824067,-0.626367,0.7448186,-1.425648,0.3524166,-0.9903384,3.339342,0.4563958,-0.2876643,1.521635,0.9508078,-0.1398541,0.3867955,-0.7550205,0.6568405,0.09419366,-1.583935,1.306094,-0.3501927,0.1794427,-0.3768163,0.9683866,-0.2442541,-1.696921,-1.8056,-0.6803037,-1.842043,0.3069353,0.9070363,-0.486526]
+spk_sig = [-1.110417,0.09703002,1.35658,0.7798632,-0.305457,-0.339204,0.6186931,
+        -0.4521213,0.3982236,-0.004530723,0.7651616,0.6500852,-0.6664245,0.1361499,
+        0.1358056,-0.2887807,-0.1280468,-0.8208137,-1.620276,-0.4628615,0.7870904,
+        -0.105754,0.9739769,-0.3258137,-0.7322628,-0.6212429,-0.5531687,-0.7796484,
+        0.7035915,1.056094,-0.4941756,-0.6521456,-0.2238328,-0.003737517,0.2165709,
+        1.200186,-0.7737719,0.492015,1.16058,0.6135428,-0.7183084,0.3153541,0.3458071,
+        -1.418189,-0.9624157,0.4168292,-1.627305,0.2742135,-0.6166027,0.1962581,
+        -0.6406527,0.4372789,-0.4296024,0.4898657,-0.9531326,-0.2945702,0.7879696,
+        -1.517101,-0.9344181,-0.5049928,-0.005040941,-0.4637912,0.8223695,-1.079849,
+        0.8871287,-0.9732434,-0.5548235,1.879138,-1.452064,-0.1975368,1.55047,
+        0.5941782,-0.52897,1.368219,0.6782904,1.202505,-0.9256122,-0.9718158,
+        -0.9570228,-0.5563112,-1.19049,-1.167985,2.606804,-2.261825,0.01340385,
+        0.2526799,-1.125458,-1.575991,-0.363153,0.3270262,1.485984,-1.769565,
+        1.541829,0.7293826,0.1743717,-0.4759418,1.523451,-2.487134,-1.824067,
+        -0.626367,0.7448186,-1.425648,0.3524166,-0.9903384,3.339342,0.4563958,
+        -0.2876643,1.521635,0.9508078,-0.1398541,0.3867955,-0.7550205,0.6568405,
+        0.09419366,-1.583935,1.306094,-0.3501927,0.1794427,-0.3768163,0.9683866,
+        -0.2442541,-1.696921,-1.8056,-0.6803037,-1.842043,0.3069353,0.9070363,-0.486526]
 
 def cosine_dist(x, y):
     nx = np.array(x)
@@ -40,16 +61,18 @@ while True:
         break
     if rec.AcceptWaveform(data):
         res = json.loads(rec.Result())
-        print ("Text:", res['text'])
-        if 'spk' in res:
-            print ("X-vector:", res['spk'])
-            print ("Speaker distance:", cosine_dist(spk_sig, res['spk']), "based on", res['spk_frames'], "frames")
+        print("Text:", res["text"])
+        if "spk" in res:
+            print("X-vector:", res["spk"])
+            print("Speaker distance:", cosine_dist(spk_sig, res["spk"]),
+                "based on", res["spk_frames"], "frames")
 
-print ("Note that second distance is not very reliable because utterance is too short. Utterances longer than 4 seconds give better xvector")
+print("Note that second distance is not very reliable because utterance is too short. "
+    "Utterances longer than 4 seconds give better xvector")
 
 res = json.loads(rec.FinalResult())
-print ("Text:", res['text'])
-if 'spk' in res:
-   print ("X-vector:", res['spk'])
-   print ("Speaker distance:", cosine_dist(spk_sig, res['spk']), "based on", res['spk_frames'], "frames")
-
+print("Text:", res["text"])
+if "spk" in res:
+    print("X-vector:", res["spk"])
+    print("Speaker distance:", cosine_dist(spk_sig, res["spk"]),
+        "based on", res["spk_frames"], "frames")

python/example/test_srt.py

@@ -1,52 +1,21 @@
 #!/usr/bin/env python3
 
-from vosk import Model, KaldiRecognizer, SetLogLevel
-import sys
-import os
-import wave
 import subprocess
-import srt
-import json
-import datetime
+import sys
+
+from vosk import Model, KaldiRecognizer, SetLogLevel
+
+SAMPLE_RATE = 16000
 
 SetLogLevel(-1)
 
-sample_rate=16000
 model = Model(lang="en-us")
-rec = KaldiRecognizer(model, sample_rate)
+rec = KaldiRecognizer(model, SAMPLE_RATE)
 rec.SetWords(True)
 
-process = subprocess.Popen(['ffmpeg', '-loglevel', 'quiet', '-i',
+with subprocess.Popen(["ffmpeg", "-loglevel", "quiet", "-i",
                             sys.argv[1],
-                            '-ar', str(sample_rate) , '-ac', '1', '-f', 's16le', '-'],
-                            stdout=subprocess.PIPE)
+                            "-ar", str(SAMPLE_RATE) , "-ac", "1", "-f", "s16le", "-"],
+                            stdout=subprocess.PIPE).stdout as stream:
 
-
-WORDS_PER_LINE = 7
-
-def transcribe():
-    results = []
-    subs = []
-    while True:
-       data = process.stdout.read(4000)
-       if len(data) == 0:
-           break
-       if rec.AcceptWaveform(data):
-           results.append(rec.Result())
-    results.append(rec.FinalResult())
-
-    for i, res in enumerate(results):
-       jres = json.loads(res)
-       if not 'result' in jres:
-           continue
-       words = jres['result']
-       for j in range(0, len(words), WORDS_PER_LINE):
-           line = words[j : j + WORDS_PER_LINE] 
-           s = srt.Subtitle(index=len(subs), 
-                   content=" ".join([l['word'] for l in line]),
-                   start=datetime.timedelta(seconds=line[0]['start']), 
-                   end=datetime.timedelta(seconds=line[-1]['end']))
-           subs.append(s)
-    return subs
-
-print (srt.compose(transcribe()))
+    print(rec.SrtResult(stream))

python/example/test_text.py

@@ -1,9 +1,9 @@
 #!/usr/bin/env python3
 
-from vosk import Model, KaldiRecognizer
 import sys
 import json
-import os
+
+from vosk import Model, KaldiRecognizer
 
 model = Model(lang="en-us")
 
@@ -13,16 +13,16 @@ rec = KaldiRecognizer(model, 16000)
 # You can also specify the possible word list
 #rec = KaldiRecognizer(model, 16000, "zero oh one two three four five six seven eight nine")
 
-wf = open(sys.argv[1], "rb")
-wf.read(44) # skip header
+with open(sys.argv[1], "rb") as wf:
+    wf.read(44) # skip header
 
-while True:
-    data = wf.read(4000)
-    if len(data) == 0:
-        break
-    if rec.AcceptWaveform(data):
-        res = json.loads(rec.Result())
-        print (res['text'])
+    while True:
+        data = wf.read(4000)
+        if len(data) == 0:
+            break
+        if rec.AcceptWaveform(data):
+            res = json.loads(rec.Result())
+            print(res["text"])
 
-res = json.loads(rec.FinalResult())
-print (res['text'])
+    res = json.loads(rec.FinalResult())
+    print(res["text"])

python/example/test_webvtt.py

@@ -1,67 +1,67 @@
 #!/usr/bin/env python3
 
-from vosk import Model, KaldiRecognizer, SetLogLevel
-from webvtt import WebVTT, Caption
 import sys
-import os
 import subprocess
 import json
 import textwrap
 
-SetLogLevel(-1)
-
-sample_rate = 16000
-model = Model(lang="en-us")
-rec = KaldiRecognizer(model, sample_rate)
-rec.SetWords(True)
+from webvtt import WebVTT, Caption
+from vosk import Model, KaldiRecognizer, SetLogLevel
 
+SAMPLE_RATE = 16000
 WORDS_PER_LINE = 7
 
+SetLogLevel(-1)
 
-def timeString(seconds):
+model = Model(lang="en-us")
+rec = KaldiRecognizer(model, SAMPLE_RATE)
+rec.SetWords(True)
+
+
+def timestring(seconds):
     minutes = seconds / 60
     seconds = seconds % 60
     hours = int(minutes / 60)
     minutes = int(minutes % 60)
-    return '%i:%02i:%06.3f' % (hours, minutes, seconds)
+    return "%i:%02i:%06.3f" % (hours, minutes, seconds)
 
 
 def transcribe():
-    command = ['ffmpeg', '-nostdin', '-loglevel', 'quiet', '-i', sys.argv[1],
-               '-ar', str(sample_rate), '-ac', '1', '-f', 's16le', '-']
-    process = subprocess.Popen(command, stdout=subprocess.PIPE)
+    command = ["ffmpeg", "-nostdin", "-loglevel", "quiet", "-i", sys.argv[1],
+               "-ar", str(SAMPLE_RATE), "-ac", "1", "-f", "s16le", "-"]
+    with subprocess.Popen(command, stdout=subprocess.PIPE) as process:
 
-    results = []
-    while True:
-        data = process.stdout.read(4000)
-        if len(data) == 0:
-            break
-        if rec.AcceptWaveform(data):
-            results.append(rec.Result())
-    results.append(rec.FinalResult())
+        results = []
+        while True:
+            data = process.stdout.read(4000)
+            if len(data) == 0:
+                break
+            if rec.AcceptWaveform(data):
+                results.append(rec.Result())
+        results.append(rec.FinalResult())
 
-    vtt = WebVTT()
-    for i, res in enumerate(results):
-        words = json.loads(res).get('result')
-        if not words:
-            continue
+        vtt = WebVTT()
+        for _, res in enumerate(results):
+            words = json.loads(res).get("result")
+            if not words:
+                continue
 
-        start = timeString(words[0]['start'])
-        end = timeString(words[-1]['end'])
-        content = ' '.join([w['word'] for w in words])
+            start = timestring(words[0]["start"])
+            end = timestring(words[-1]["end"])
+            content = " ".join([w["word"] for w in words])
 
-        caption = Caption(start, end, textwrap.fill(content))
-        vtt.captions.append(caption)
+            caption = Caption(start, end, textwrap.fill(content))
+            vtt.captions.append(caption)
 
-    # save or return webvtt
-    if len(sys.argv) > 2:
-        vtt.save(sys.argv[2])
-    else:
-        print(vtt.content)
+        # save or return webvtt
+        if len(sys.argv) > 2:
+            vtt.save(sys.argv[2])
+        else:
+            print(vtt.content)
 
 
-if __name__ == '__main__':
-    if not (1 < len(sys.argv) < 4):
-        print(f'Usage: {sys.argv[0]} audiofile [output file]')
-        exit(1)
+if __name__ == "__main__":
+    if not 1 < len(sys.argv) < 4:
+        print("Usage: {} audiofile [output file]".format(sys.argv[0]))
+        sys.exit(1)
     transcribe()

python/example/test_words.py

@@ -1,19 +1,22 @@
 #!/usr/bin/env python3
 
-from vosk import Model, KaldiRecognizer
-import sys
-import os
 import wave
+import sys
+
+from vosk import Model, KaldiRecognizer
 
 wf = wave.open(sys.argv[1], "rb")
 if wf.getnchannels() != 1 or wf.getsampwidth() != 2 or wf.getcomptype() != "NONE":
-    print ("Audio file must be WAV format mono PCM.")
-    exit (1)
+    print("Audio file must be WAV format mono PCM.")
+    sys.exit(1)
 
 model = Model(lang="en-us")
 
-# You can also specify the possible word or phrase list as JSON list, the order doesn't have to be strict
-rec = KaldiRecognizer(model, wf.getframerate(), '["oh one two three four five six seven eight nine zero", "[unk]"]')
+# You can also specify the possible word or phrase list as JSON list,
+# the order doesn't have to be strict
+rec = KaldiRecognizer(model,
+    wf.getframerate(),
+    '["oh one two three", "four five six", "seven eight nine zero", "[unk]"]')
 
 while True:
     data = wf.readframes(4000)
@@ -21,6 +24,7 @@ while True:
         break
     if rec.AcceptWaveform(data):
         print(rec.Result())
+        rec.SetGrammar('["one zero one two three oh", "four five six", "seven eight nine zero", "[unk]"]')
     else:
         print(rec.PartialResult())
 

python/setup.py

@@ -1,14 +1,15 @@
 import os
-import sys
 import setuptools
 import shutil
 import glob
 import platform
 
 # Figure out environment for cross-compile
-vosk_source = os.getenv("VOSK_SOURCE", os.path.abspath(os.path.join(os.path.dirname(__file__), "..")))
-system = os.environ.get('VOSK_PLATFORM', platform.system())
+vosk_source = os.getenv("VOSK_SOURCE", os.path.abspath(os.path.join(os.path.dirname(__file__),
+    "..")))
+system = os.environ.get('VOSK_SYSTEM', platform.system())
 architecture = os.environ.get('VOSK_ARCHITECTURE', platform.architecture()[0])
+machine = os.environ.get('VOSK_MACHINE', platform.machine())
 
 # Copy precompmilled libraries
 for lib in glob.glob(os.path.join(vosk_source, "src/lib*.*")):
@@ -30,21 +31,21 @@ else:
                 oses = 'win32'
             elif system == 'Windows' and architecture == '64bit':
                 oses = 'win_amd64'
-            elif system == 'Linux' and architecture == '64bit':
-                oses = 'linux_x86_64'
+            elif system == 'Linux' and machine == 'aarch64' and architecture == '64bit':
+                oses = 'manylinux2014_aarch64'
             elif system == 'Linux':
-                oses = 'linux_' + architecture
+                oses = 'linux_' + machine
             else:
                 raise TypeError("Unknown build environment")
             return 'py3', abi, oses
     cmdclass = {'bdist_wheel': bdist_wheel_tag_name}
 
-with open("README.md", "r") as fh:
-    long_description = fh.read()
+with open("README.md", "rb") as fh:
+    long_description = fh.read().decode("utf-8")
 
 setuptools.setup(
     name="vosk",
-    version="0.3.41",
+    version="0.3.45",
     author="Alpha Cephei Inc",
     author_email="contact@alphacephei.com",
     description="Offline open source speech recognition API based on Kaldi and Vosk",
@@ -68,7 +69,7 @@ setuptools.setup(
     cmdclass=cmdclass,
     python_requires='>=3',
     zip_safe=False, # Since we load so file from the filesystem, we can not run from zip file
-    setup_requires=['cffi>=1.0', 'requests', 'tqdm'],
-    install_requires=['cffi>=1.0', 'requests', 'tqdm'],
+    setup_requires=['cffi>=1.0', 'requests', 'tqdm', 'srt', 'websockets'],
+    install_requires=['cffi>=1.0', 'requests', 'tqdm', 'srt', 'websockets'],
     cffi_modules=['vosk_builder.py:ffibuilder'],
 )

python/test/transcribe_scp.py

@@ -1,14 +1,13 @@
 #!/usr/bin/env python3
 
+import wave
+import json
+import sys
+
 from multiprocessing.dummy import Pool
 from vosk import Model, KaldiRecognizer
 
-import sys
-import os
-import wave
-import json
-
-model = Model("model")
+model = Model("en-us")
 
 def recognize(line):
     uid, fn = line.split()
@@ -22,14 +21,14 @@ def recognize(line):
             break
         if rec.AcceptWaveform(data):
             jres = json.loads(rec.Result())
-            text = text + " " + jres['text']
+            text = text + " " + jres["text"]
     jres = json.loads(rec.FinalResult())
-    text = text + " " + jres['text']
-    return (uid + text)
+    text = text + " " + jres["text"]
+    return uid + text
 
 def main():
     p = Pool(8)
-    texts = p.map(recognize, open(sys.argv[1]).readlines())
+    texts = p.map(recognize, open(sys.argv[1], encoding="uft-8").readlines())
     print ("\n".join(texts))
 
 main()

python/vosk/__init__.py

@@ -1,5 +1,8 @@
 import os
 import sys
+import srt
+import datetime
+import json
 
 import requests
 from urllib.request import urlretrieve
@@ -10,21 +13,22 @@ from .vosk_cffi import ffi as _ffi
 from tqdm import tqdm
 
 # Remote location of the models and local folders
-MODEL_PRE_URL = 'https://alphacephei.com/vosk/models/'
-MODEL_LIST_URL = MODEL_PRE_URL + 'model-list.json'
-MODEL_DIRS = [os.getenv('VOSK_MODEL_PATH'), Path('/usr/share/vosk'), Path.home() / 'AppData/Local/vosk', Path.home() / '.cache/vosk']
+MODEL_PRE_URL = "https://alphacephei.com/vosk/models/"
+MODEL_LIST_URL = MODEL_PRE_URL + "model-list.json"
+MODEL_DIRS = [os.getenv("VOSK_MODEL_PATH"), Path("/usr/share/vosk"),
+        Path.home() / "AppData/Local/vosk", Path.home() / ".cache/vosk"]
 
 def open_dll():
     dlldir = os.path.abspath(os.path.dirname(__file__))
-    if sys.platform == 'win32':
+    if sys.platform == "win32":
         # We want to load dependencies too
-        os.environ["PATH"] = dlldir + os.pathsep + os.environ['PATH']
-        if hasattr(os, 'add_dll_directory'):
+        os.environ["PATH"] = dlldir + os.pathsep + os.environ["PATH"]
+        if hasattr(os, "add_dll_directory"):
             os.add_dll_directory(dlldir)
         return _ffi.dlopen(os.path.join(dlldir, "libvosk.dll"))
-    elif sys.platform == 'linux':
+    elif sys.platform == "linux":
         return _ffi.dlopen(os.path.join(dlldir, "libvosk.so"))
-    elif sys.platform == 'darwin':
+    elif sys.platform == "darwin":
         return _ffi.dlopen(os.path.join(dlldir, "libvosk.dyld"))
     else:
         raise TypeError("Unsupported platform")
@@ -32,23 +36,23 @@ def open_dll():
 _c = open_dll()
 
 def list_models():
-    response = requests.get(MODEL_LIST_URL)
+    response = requests.get(MODEL_LIST_URL, timeout=10)
     for model in response.json():
-        print(model['name']) 
+        print(model["name"])
 
 def list_languages():
-    response = requests.get(MODEL_LIST_URL)
-    languages = set([m['lang'] for m in response.json()])
+    response = requests.get(MODEL_LIST_URL, timeout=10)
+    languages = {m["lang"] for m in response.json()}
     for lang in languages:
         print (lang)
 
-class Model(object):
+class Model:
     def __init__(self, model_path=None, model_name=None, lang=None):
-        if model_path != None:
-            self._handle = _c.vosk_model_new(model_path.encode('utf-8'))
+        if model_path is not None:
+            self._handle = _c.vosk_model_new(model_path.encode("utf-8"))
         else:
             model_path = self.get_model_path(model_name, lang)
-            self._handle = _c.vosk_model_new(model_path.encode('utf-8'))
+            self._handle = _c.vosk_model_new(model_path.encode("utf-8"))
         if self._handle == _ffi.NULL:
             raise Exception("Failed to create a model")
 
@@ -56,7 +60,7 @@ class Model(object):
         _c.vosk_model_free(self._handle)
 
     def vosk_model_find_word(self, word):
-        return _c.vosk_model_find_word(self._handle, word.encode('utf-8'))
+        return _c.vosk_model_find_word(self._handle, word.encode("utf-8"))
 
     def get_model_path(self, model_name, lang):
         if model_name is None:
@@ -73,10 +77,11 @@ class Model(object):
             model_file = [model for model in model_file_list if model == model_name]
             if model_file != []:
                 return Path(directory, model_file[0])
-        response = requests.get(MODEL_LIST_URL)
-        result_model = [model['name'] for model in response.json() if model['name'] == model_name]
+        response = requests.get(MODEL_LIST_URL, timeout=10)
+        result_model = [model["name"] for model in response.json() if model["name"] == model_name]
         if result_model == []:
-            raise Exception("model name %s does not exist" % (model_name))
+            print("model name %s does not exist" % (model_name))
+            sys.exit(1)
         else:
             self.download_model(Path(directory, result_model[0]))
             return Path(directory, result_model[0])
@@ -86,29 +91,33 @@ class Model(object):
             if directory is None or not Path(directory).exists():
                 continue
             model_file_list = os.listdir(directory)
-            model_file = [model for model in model_file_list if match(f"vosk-model(-small)?-{lang}", model)]
+            model_file = [model for model in model_file_list if
+                    match(r"vosk-model(-small)?-{}".format(lang), model)]
             if model_file != []:
                 return Path(directory, model_file[0])
-        response = requests.get(MODEL_LIST_URL)
-        result_model = [model['name'] for model in response.json() if model['lang'] == lang and model['type'] == 'small' and model['obsolete'] == 'false']
+        response = requests.get(MODEL_LIST_URL, timeout=10)
+        result_model = [model["name"] for model in response.json() if
+                model["lang"] == lang and model["type"] == "small" and model["obsolete"] == "false"]
         if result_model == []:
-            raise Exception("lang %s does not exist" % (lang))
+            print("lang %s does not exist" % (lang))
+            sys.exit(1)
         else:
             self.download_model(Path(directory, result_model[0]))
             return Path(directory, result_model[0])
 
     def download_model(self, model_name):
-        if not MODEL_DIRS[3].exists():
-            MODEL_DIRS[3].mkdir()
-        with tqdm(unit='B', unit_scale=True, unit_divisor=1024, miniters=1,
-                desc=(MODEL_PRE_URL + str(model_name.name) + '.zip').split('/')[-1]) as t:
+        if not (model_name.parent).exists():
+            (model_name.parent).mkdir(parents=True)
+        with tqdm(unit="B", unit_scale=True, unit_divisor=1024, miniters=1,
+                desc=(MODEL_PRE_URL + str(model_name.name) + ".zip").rsplit("/",
+                    maxsplit=1)[-1]) as t:
             reporthook = self.download_progress_hook(t)
-            urlretrieve(MODEL_PRE_URL + str(model_name.name) + '.zip', str(model_name) + '.zip', 
-                reporthook=reporthook, data=None)
+            urlretrieve(MODEL_PRE_URL + str(model_name.name) + ".zip",
+                    str(model_name) + ".zip", reporthook=reporthook, data=None)
             t.total = t.n
-            with ZipFile(str(model_name) + '.zip', 'r') as model_ref:
+            with ZipFile(str(model_name) + ".zip", "r") as model_ref:
                 model_ref.extractall(model_name.parent)
-            Path(str(model_name) + '.zip').unlink()
+            Path(str(model_name) + ".zip").unlink()
 
     def download_progress_hook(self, t):
         last_b = [0]
@@ -120,10 +129,10 @@ class Model(object):
             return displayed
         return update_to
 
-class SpkModel(object):
+class SpkModel:
 
     def __init__(self, model_path):
-        self._handle = _c.vosk_spk_model_new(model_path.encode('utf-8'))
+        self._handle = _c.vosk_spk_model_new(model_path.encode("utf-8"))
 
         if self._handle == _ffi.NULL:
             raise Exception("Failed to create a speaker model")
@@ -131,15 +140,17 @@ class SpkModel(object):
     def __del__(self):
         _c.vosk_spk_model_free(self._handle)
 
-class KaldiRecognizer(object):
+class KaldiRecognizer:
 
     def __init__(self, *args):
         if len(args) == 2:
             self._handle = _c.vosk_recognizer_new(args[0]._handle, args[1])
-        elif len(args) == 3 and type(args[2]) is SpkModel:
-            self._handle = _c.vosk_recognizer_new_spk(args[0]._handle, args[1], args[2]._handle)
-        elif len(args) == 3 and type(args[2]) is str:
-            self._handle = _c.vosk_recognizer_new_grm(args[0]._handle, args[1], args[2].encode('utf-8'))
+        elif len(args) == 3 and isinstance(args[2], SpkModel):
+            self._handle = _c.vosk_recognizer_new_spk(args[0]._handle,
+                    args[1], args[2]._handle)
+        elif len(args) == 3 and isinstance(args[2], str):
+            self._handle = _c.vosk_recognizer_new_grm(args[0]._handle,
+                    args[1], args[2].encode("utf-8"))
         else:
             raise TypeError("Unknown arguments")
 
@@ -164,6 +175,9 @@ class KaldiRecognizer(object):
     def SetSpkModel(self, spk_model):
         _c.vosk_recognizer_set_spk_model(self._handle, spk_model._handle)
 
+    def SetGrammar(self, grammar):
+        _c.vosk_recognizer_set_grm(self._handle, grammar.encode("utf-8"))
+
     def AcceptWaveform(self, data):
         res = _c.vosk_recognizer_accept_waveform(self._handle, data, len(data))
         if res < 0:
@@ -171,17 +185,43 @@ class KaldiRecognizer(object):
         return res
 
     def Result(self):
-        return _ffi.string(_c.vosk_recognizer_result(self._handle)).decode('utf-8')
+        return _ffi.string(_c.vosk_recognizer_result(self._handle)).decode("utf-8")
 
     def PartialResult(self):
-        return _ffi.string(_c.vosk_recognizer_partial_result(self._handle)).decode('utf-8')
+        return _ffi.string(_c.vosk_recognizer_partial_result(self._handle)).decode("utf-8")
 
     def FinalResult(self):
-        return _ffi.string(_c.vosk_recognizer_final_result(self._handle)).decode('utf-8')
+        return _ffi.string(_c.vosk_recognizer_final_result(self._handle)).decode("utf-8")
 
     def Reset(self):
         return _c.vosk_recognizer_reset(self._handle)
 
+    def SrtResult(self, stream, words_per_line = 7):
+        results = []
+
+        while True:
+            data = stream.read(4000)
+            if len(data) == 0:
+                break
+            if self.AcceptWaveform(data):
+                results.append(self.Result())
+        results.append(self.FinalResult())
+
+        subs = []
+        for res in results:
+            jres = json.loads(res)
+            if not "result" in jres:
+                continue
+            words = jres["result"]
+            for j in range(0, len(words), words_per_line):
+                line = words[j : j + words_per_line]
+                s = srt.Subtitle(index=len(subs),
+                        content=" ".join([l["word"] for l in line]),
+                        start=datetime.timedelta(seconds=line[0]["start"]),
+                        end=datetime.timedelta(seconds=line[-1]["end"]))
+                subs.append(s)
+
+        return srt.compose(subs)
 
 def SetLogLevel(level):
     return _c.vosk_set_log_level(level)
@@ -194,10 +234,10 @@ def GpuInit():
 def GpuThreadInit():
     _c.vosk_gpu_thread_init()
 
-class BatchModel(object):
+class BatchModel:
 
-    def __init__(self, *args):
-        self._handle = _c.vosk_batch_model_new()
+    def __init__(self, model_path, *args):
+        self._handle = _c.vosk_batch_model_new(model_path.encode('utf-8'))
 
         if self._handle == _ffi.NULL:
             raise Exception("Failed to create a model")
@@ -208,7 +248,7 @@ class BatchModel(object):
     def Wait(self):
         _c.vosk_batch_model_wait(self._handle)
 
-class BatchRecognizer(object):
+class BatchRecognizer:
 
     def __init__(self, *args):
         self._handle = _c.vosk_batch_recognizer_new(args[0]._handle, args[1])
@@ -224,7 +264,7 @@ class BatchRecognizer(object):
 
     def Result(self):
         ptr = _c.vosk_batch_recognizer_front_result(self._handle)
-        res = _ffi.string(ptr).decode('utf-8')
+        res = _ffi.string(ptr).decode("utf-8")
         _c.vosk_batch_recognizer_pop(self._handle)
         return res
 

python/vosk/transcriber/cli.py

@@ -1,41 +1,49 @@
 #!/usr/bin/env python3
 
-import logging
 import argparse
+import logging
+import sys
+import os
 
 from pathlib import Path
 from vosk import list_models, list_languages
 from vosk.transcriber.transcriber import Transcriber
 
 parser = argparse.ArgumentParser(
-        description = 'Transcribe audio file and save result in selected format')
+        description = "Transcribe audio file and save result in selected format")
 parser.add_argument(
-        '--model', '-m', type=str,
-        help='model path')
+        "--model", "-m", type=str,
+        help="model path")
 parser.add_argument(
-        '--list-models', default=False, action='store_true', 
-        help='list available models')
+        "--server", "-s", const="ws://localhost:2700", action="store_const",
+        help="use server for recognition")
 parser.add_argument(
-        '--list-languages', default=False, action='store_true',
-        help='list available languages')
+        "--list-models", default=False, action="store_true",
+        help="list available models")
 parser.add_argument(
-        '--model-name', '-n', type=str,
-        help='select model by name')
+        "--list-languages", default=False, action="store_true",
+        help="list available languages")
 parser.add_argument(
-        '--lang', '-l', default='en-us', type=str,
-        help='select model by language')
+        "--model-name", "-n", type=str,
+        help="select model by name")
 parser.add_argument(
-        '--input', '-i', type=str,
-        help='audiofile')
+        "--lang", "-l", default="en-us", type=str,
+        help="select model by language")
 parser.add_argument(
-        '--output', '-o', default='', type=str,
-        help='optional output filename path')
+        "--input", "-i", type=str,
+        help="audiofile")
 parser.add_argument(
-        '--output-type', '-t', default='txt', type=str,
-        help='optional arg output data type')
+        "--output", "-o", default="", type=str,
+        help="optional output filename path")
 parser.add_argument(
-        '--log-level', default='INFO',
-        help='logging level')
+        "--output-type", "-t", default="txt", type=str,
+        help="optional arg output data type")
+parser.add_argument(
+        "--tasks", "-ts", default=10, type=int,
+        help="number of parallel recognition tasks")
+parser.add_argument(
+        "--log-level", default="INFO",
+        help="logging level")
 
 def main():
 
@@ -43,36 +51,39 @@ def main():
     log_level = args.log_level.upper()
     logging.getLogger().setLevel(log_level)
 
-    if args.list_models == True:
+    if args.list_models is True:
         list_models()
         return
 
-    if args.list_languages == True:
+    if args.list_languages is True:
         list_languages()
         return
 
     if not args.input:
-        logging.info('Please specify input file or directory')
-        exit(1)
+        logging.info("Please specify input file or directory")
+        sys.exit(1)
 
     if not Path(args.input).exists():
-        logging.info('File %s does not exist, please specify an existing file/directory' % (args.input))
-        exit(1)
-
-    if args.output !='' and not Path(args.output).exists():
-        logging.info('Output %s does not exist, please specify an existing file' % (args.output))
-        exit(1)
+        logging.info("File/folder {args.input} does not exist, "\
+            "please specify an existing file/directory")
+        sys.exit(1)
 
     transcriber = Transcriber(args)
 
-    if Path(args.input).is_dir() and Path(args.output).is_dir():
-        transcriber.process_dir(args)
-        return
-    elif Path(args.input).is_file() and (args.output=='' or Path(args.output).is_file()):
-        transcriber.process_file(args)
+    if Path(args.input).is_dir():
+        task_list = [(Path(args.input, fn),
+            Path(args.output,
+            Path(fn).stem).with_suffix("." + args.output_type)) for fn in os.listdir(args.input)]
+    elif Path(args.input).is_file():
+        if args.output == "":
+            task_list = [(Path(args.input), args.output)]
+        else:
+            task_list = [(Path(args.input), Path(args.output))]
     else:
-        logging.info('Wrong arguments, input and output must be same type')
-        exit(1)
+        logging.info("Wrong arguments")
+        sys.exit(1)
+
+    transcriber.process_task_list(task_list)
 
 if __name__ == "__main__":
     main()

python/vosk/transcriber/transcriber.py

@@ -1,89 +1,195 @@
 import json
-import subprocess
+import logging
+import asyncio
+import websockets
 import srt
 import datetime
-import os
-import logging
+import shlex
+import subprocess
 
-from pathlib import Path
-from timeit import default_timer as timer
 from vosk import KaldiRecognizer, Model
+from queue import Queue
+from timeit import default_timer as timer
+from multiprocessing.dummy import Pool
+
+CHUNK_SIZE = 4000
+SAMPLE_RATE = 16000.0
 
 class Transcriber:
 
     def __init__(self, args):
         self.model = Model(model_path=args.model, model_name=args.model_name, lang=args.lang)
         self.args = args
+        self.queue = Queue()
 
     def recognize_stream(self, rec, stream):
         tot_samples = 0
         result = []
+
         while True:
-            data = stream.stdout.read(4000)
+            data = stream.stdout.read(CHUNK_SIZE)
+
             if len(data) == 0:
                 break
+
+            tot_samples += len(data)
             if rec.AcceptWaveform(data):
-                tot_samples += len(data)
-                result.append(json.loads(rec.Result())) 
-        result.append(json.loads(rec.FinalResult()))
+                jres = json.loads(rec.Result())
+                logging.info(jres)
+                result.append(jres)
+            else:
+                jres = json.loads(rec.PartialResult())
+                if jres["partial"] != "":
+                    logging.info(jres)
+
+        jres = json.loads(rec.FinalResult())
+        result.append(jres)
+
         return result, tot_samples
 
+    async def recognize_stream_server(self, proc):
+        async with websockets.connect(self.args.server) as websocket:
+            tot_samples = 0
+            result = []
+
+            await websocket.send('{ "config" : { "sample_rate" : %f } }' % (SAMPLE_RATE))
+            while True:
+                data = await proc.stdout.read(CHUNK_SIZE)
+                tot_samples += len(data)
+                if len(data) == 0:
+                    break
+                await websocket.send(data)
+                jres = json.loads(await websocket.recv())
+                logging.info(jres)
+                if not "partial" in jres:
+                    result.append(jres)
+            await websocket.send('{"eof" : 1}')
+            jres = json.loads(await websocket.recv())
+            logging.info(jres)
+            result.append(jres)
+
+            return result, tot_samples
+
+
     def format_result(self, result, words_per_line=7):
-        final_result = ''
-        if self.args.output_type == 'srt':
+        processed_result = ""
+        if self.args.output_type == "srt":
             subs = []
-            for i, res in enumerate(result):
-                if not 'result' in res:
+
+            for _, res in enumerate(result):
+                if not "result" in res:
                     continue
-                words = res['result']
+                words = res["result"]
+
                 for j in range(0, len(words), words_per_line):
                     line = words[j : j + words_per_line]
                     s = srt.Subtitle(index=len(subs),
-                            content = ' '.join([l['word'] for l in line]),
-                            start=datetime.timedelta(seconds=line[0]['start']),
-                            end=datetime.timedelta(seconds=line[-1]['end']))
+                            content = " ".join([l["word"] for l in line]),
+                            start=datetime.timedelta(seconds=line[0]["start"]),
+                            end=datetime.timedelta(seconds=line[-1]["end"]))
                     subs.append(s)
-            final_result = srt.compose(subs)
-        elif self.args.output_type == 'txt':
-            for part in result:
-                final_result += part['text'] + ' '
-        return final_result
+            processed_result = srt.compose(subs)
 
+        elif self.args.output_type == "txt":
+            for part in result:
+                if part["text"] != "":
+                    processed_result += part["text"] + "\n"
+
+        elif self.args.output_type == "json":
+            monologues = {"schemaVersion":"2.0", "monologues":[], "text":[]}
+            for part in result:
+                if part["text"] != "":
+                    monologue["text"] += part["text"]
+            for _, res in enumerate(result):
+                if not "result" in res:
+                    continue
+                monologue = { "speaker": {"id": "unknown", "name": None}, "start": 0, "end": 0, "terms": []}
+                monologue["start"] = res["result"][0]["start"]
+                monologue["end"] = res["result"][-1]["end"]
+                monologue["terms"] = [{"confidence": t["conf"], "start": t["start"], "end": t["end"], "text": t["word"], "type": "WORD" } for t in res["result"]]
+                monologues["monologues"].append(monologue)
+            processed_result = json.dumps(monologues)
+        return processed_result
 
     def resample_ffmpeg(self, infile):
-        stream = subprocess.Popen(
-            ['ffmpeg', '-nostdin', '-loglevel', 'quiet', '-i', 
-            infile, 
-            '-ar', '16000','-ac', '1', '-f', 's16le', '-'], 
-            stdout=subprocess.PIPE)
+        cmd = shlex.split("ffmpeg -nostdin -loglevel quiet "
+                "-i \'{}\' -ar {} -ac 1 -f s16le -".format(str(infile), SAMPLE_RATE))
+        stream = subprocess.Popen(cmd, stdout=subprocess.PIPE)
         return stream
 
+    async def resample_ffmpeg_async(self, infile):
+        cmd = "ffmpeg -nostdin -loglevel quiet "\
+        "-i \'{}\' -ar {} -ac 1 -f s16le -".format(str(infile), SAMPLE_RATE)
+        return await asyncio.create_subprocess_shell(cmd, stdout=subprocess.PIPE)
 
-    def process_entry(self, inputdata):
-        logging.info(f'Recognizing {inputdata[0]}')
+    async def server_worker(self):
+        while True:
+            try:
+                input_file, output_file = self.queue.get_nowait()
+            except Exception:
+                break
 
-        rec = KaldiRecognizer(self.model, 16000)
-        rec.SetWords(True)
+            logging.info("Recognizing {}".format(input_file))
+            start_time = timer()
+            proc = await self.resample_ffmpeg_async(input_file)
+            result, tot_samples = await self.recognize_stream_server(proc)
 
-        stream = self.resample_ffmpeg(inputdata[0])
-        result, tot_samples = self.recognize_stream(rec, stream)
-        final_result = self.format_result(result)
+            processed_result = self.format_result(result)
+            if output_file != "":
+                logging.info("File {} processing complete".format(output_file))
+                with open(output_file, "w", encoding="utf-8") as fh:
+                    fh.write(processed_result)
+            else:
+                print(processed_result)
 
-        if inputdata[1] != '':
-            with open(inputdata[1], 'w', encoding='utf-8') as fh:
-                fh.write(final_result)
-        else:
-            print(final_result)
-        return final_result, tot_samples
+            await proc.wait()
 
+            elapsed = timer() - start_time
+            logging.info("Execution time: {:.3f} sec; "\
+                    "xRT {:.3f}".format(elapsed, float(elapsed) * (2 * SAMPLE_RATE) / tot_samples))
+            self.queue.task_done()
 
-    def process_directory(self,args):
-        task_list = [(Path(args.input, fn), Path(args.output, Path(fn).stem).with_suffix('.' + args.output_type)) for fn in os.listdir(args.input)]
-        with Pool() as pool:
-            pool.map(self.process_entry, file_list)
-
-    def process_file(self, args):
+    def pool_worker(self, inputdata):
+        logging.info("Recognizing {}".format(inputdata[0]))
         start_time = timer()
-        final_result, tot_samples = self.process_entry([args.input, args.output])
+
+        try:
+            stream = self.resample_ffmpeg(inputdata[0])
+        except FileNotFoundError as e:
+            print(e, "Missing FFMPEG, please install and try again")
+            return
+        except Exception as e:
+            logging.info(e)
+            return
+
+        rec = KaldiRecognizer(self.model, SAMPLE_RATE)
+        rec.SetWords(True)
+        result, tot_samples = self.recognize_stream(rec, stream)
+        processed_result = self.format_result(result)
+
+        if inputdata[1] != "":
+            logging.info("File {} processing complete".format(inputdata[1]))
+            with open(inputdata[1], "w", encoding="utf-8") as fh:
+                fh.write(processed_result)
+        else:
+            print(processed_result)
+
         elapsed = timer() - start_time
-        logging.info(f'''Execution time: {elapsed:.3f} sec; xRT: {format(tot_samples / 16000.0 / float(elapsed), '.3f')}''')
+        logging.info("Execution time: {:.3f} sec; "\
+                "xRT {:.3f}".format(elapsed, float(elapsed) * (2 * SAMPLE_RATE) / tot_samples))
+
+    async def process_task_list_server(self, task_list):
+        for x in task_list:
+            self.queue.put(x)
+        workers = [asyncio.create_task(self.server_worker()) for i in range(self.args.tasks)]
+        await asyncio.gather(*workers)
+
+    def process_task_list_pool(self, task_list):
+        with Pool() as pool:
+            pool.map(self.pool_worker, task_list)
+
+    def process_task_list(self, task_list):
+        if self.args.server is None:
+            self.process_task_list_pool(task_list)
+        else:
+            asyncio.run(self.process_task_list_server(task_list))

ruby/lib/vosk.rb

@@ -0,0 +1,5 @@
+class Vosk
+  def self.hi
+    puts "Hello world!"
+  end
+end

ruby/vosk.gemspec

@@ -0,0 +1,12 @@
+Gem::Specification.new do |s|
+  s.name        = "vosk"
+  s.version     = "0.3.45"
+  s.summary     = "Offline speech recognition API"
+  s.description = "Vosk is an offline open source speech recognition toolkit. It enables speech recognition for 20+ languages and dialects - English, Indian English, German, French, Spanish, Portuguese, Chinese, Russian, Turkish, Vietnamese, Italian, Dutch, Catalan, Arabic, Greek, Farsi, Filipino, Ukrainian, Kazakh, Swedish, Japanese, Esperanto, Hindi, Czech, Polish. More to come."
+  s.authors     = ["Alpha Cephei Inc"]
+  s.email       = "contact@alphacphei.com"
+  s.files       = ["lib/vosk.rb"]
+  s.homepage    =
+    "https://rubygems.org/gems/vosk"
+  s.license       = "Apache 2.0"
+end

rust/README.md

@@ -0,0 +1,5 @@
+See
+
+https://github.com/Bear-03/vosk-rs
+
+https://crates.io/crates/vosk

src/batch_model.cc

@@ -20,21 +20,21 @@ using namespace fst;
 using namespace kaldi::nnet3;
 using CorrelationID = CudaOnlinePipelineDynamicBatcher::CorrelationID;
 
-BatchModel::BatchModel() {
+BatchModel::BatchModel(const char *model_path) : model_path_str_(model_path) {
     BatchedThreadedNnet3CudaOnlinePipelineConfig batched_decoder_config;
 
     kaldi::ParseOptions po("something");
     batched_decoder_config.Register(&po);
-    po.ReadConfigFile("model/conf/model.conf");
+    po.ReadConfigFile(model_path_str_ + "/conf/model.conf");
 
     struct stat buffer;
 
-    string nnet3_rxfilename_ = "model/am/final.mdl";
-    string hclg_fst_rxfilename_ = "model/graph/HCLG.fst";
-    string word_syms_rxfilename_ = "model/graph/words.txt";
-    string winfo_rxfilename_ = "model/graph/phones/word_boundary.int";
-    string std_fst_rxfilename_ = "model/rescore/G.fst";
-    string carpa_rxfilename_ = "model/rescore/G.carpa";
+    string nnet3_rxfilename_ = model_path_str_ + "/am/final.mdl";
+    string hclg_fst_rxfilename_ = model_path_str_ + "/graph/HCLG.fst";
+    string word_syms_rxfilename_ = model_path_str_ + "/graph/words.txt";
+    string winfo_rxfilename_ = model_path_str_ + "/graph/phones/word_boundary.int";
+    string std_fst_rxfilename_ = model_path_str_ + "/rescore/G.fst";
+    string carpa_rxfilename_ = model_path_str_ + "/rescore/G.carpa";
 
     trans_model_ = new kaldi::TransitionModel();
     nnet_ = new kaldi::nnet3::AmNnetSimple();
@@ -72,9 +72,23 @@ BatchModel::BatchModel() {
     batched_decoder_config.reset_on_endpoint = true;
     batched_decoder_config.use_gpu_feature_extraction = true;
 
-    batched_decoder_config.feature_opts.feature_type = "mfcc";
-    batched_decoder_config.feature_opts.mfcc_config = "model/conf/mfcc.conf";
-    batched_decoder_config.feature_opts.ivector_extraction_config = "model/conf/ivector.conf";
+    feature_info_.feature_type = "mfcc";
+    ReadConfigFromFile(model_path_str_ + "/conf/mfcc.conf", &feature_info_.mfcc_opts);
+    feature_info_.silence_weighting_config.silence_weight = 1e-3;
+    feature_info_.silence_weighting_config.silence_phones_str = batched_decoder_config.decoder_opts.endpointing_config.silence_phones;
+
+    OnlineIvectorExtractionConfig ivector_extraction_opts;
+    ivector_extraction_opts.splice_config_rxfilename = model_path_str_ + "/ivector/splice.conf";
+    ivector_extraction_opts.cmvn_config_rxfilename = model_path_str_ + "/ivector/online_cmvn.conf";
+    ivector_extraction_opts.lda_mat_rxfilename = model_path_str_ + "/ivector/final.mat";
+    ivector_extraction_opts.global_cmvn_stats_rxfilename = model_path_str_ + "/ivector/global_cmvn.stats";
+    ivector_extraction_opts.diag_ubm_rxfilename = model_path_str_ + "/ivector/final.dubm";
+    ivector_extraction_opts.ivector_extractor_rxfilename = model_path_str_ + "/ivector/final.ie";
+    ivector_extraction_opts.max_count = 100;
+    feature_info_.use_ivectors = true;
+    feature_info_.ivector_extractor_info.Init(ivector_extraction_opts);
+
+
     batched_decoder_config.decoder_opts.max_active = 7000;
     batched_decoder_config.decoder_opts.default_beam = 13.0;
     batched_decoder_config.decoder_opts.lattice_beam = 6.0;
@@ -88,7 +102,7 @@ BatchModel::BatchModel() {
     batched_decoder_config.compute_opts.frames_per_chunk = std::max(51, (nnet_right_context + 3 - nnet_right_context % 3));
 
     cuda_pipeline_ = new BatchedThreadedNnet3CudaOnlinePipeline 
-         (batched_decoder_config, *hclg_fst_, *nnet_, *trans_model_);
+         (batched_decoder_config, feature_info_, *hclg_fst_, *nnet_, *trans_model_);
     cuda_pipeline_->SetSymbolTable(*word_syms_);
 
     CudaOnlinePipelineDynamicBatcherConfig dynamic_batcher_config;

src/batch_model.h

@@ -42,7 +42,7 @@ class BatchRecognizer;
 
 class BatchModel {
     public:
-        BatchModel();
+        BatchModel(const char *model_path);
         ~BatchModel();
 
         uint64_t GetID(BatchRecognizer *recognizer);
@@ -51,6 +51,9 @@ class BatchModel {
     private:
         friend class BatchRecognizer;
 
+        std::string model_path_str_;
+
+        OnlineNnet2FeaturePipelineInfo feature_info_;
         kaldi::TransitionModel *trans_model_ = nullptr;
         kaldi::nnet3::AmNnetSimple *nnet_ = nullptr;
         const fst::SymbolTable *word_syms_ = nullptr;

src/model.cc

@@ -282,7 +282,10 @@ void Model::ReadDataFiles()
         KALDI_LOG << "Loading HCL and G from " << hcl_fst_rxfilename_ << " " << g_fst_rxfilename_;
         hcl_fst_ = fst::StdFst::Read(hcl_fst_rxfilename_);
         g_fst_ = fst::StdFst::Read(g_fst_rxfilename_);
-        ReadIntegerVectorSimple(disambig_rxfilename_, &disambig_);
+        if (!ReadIntegerVectorSimple(disambig_rxfilename_, &disambig_)) {
+            KALDI_ERR << "Could not read disambig symbol table from file "
+                      << disambig_rxfilename_;
+        }
     }
 
     if (hclg_fst_ && hclg_fst_->OutputSymbols()) {
@@ -297,7 +300,9 @@ void Model::ReadDataFiles()
                       << word_syms_rxfilename_;
         word_syms_loaded_ = word_syms_;
     }
-    KALDI_ASSERT(word_syms_);
+    if (!word_syms_) {
+        KALDI_ERR << "Word symbol table empty";
+    }
 
     if (stat(winfo_rxfilename_.c_str(), &buffer) == 0) {
         KALDI_LOG << "Loading winfo " << winfo_rxfilename_;

src/recognizer.cc

@@ -54,45 +54,7 @@ Recognizer::Recognizer(Model *model, float sample_frequency, char const *grammar
     silence_weighting_ = new kaldi::OnlineSilenceWeighting(*model_->trans_model_, model_->feature_info_.silence_weighting_config, 3);
 
     if (model_->hcl_fst_) {
-        json::JSON obj;
-        obj = json::JSON::Load(grammar);
-
-        if (obj.length() <= 0) {
-            KALDI_WARN << "Expecting array of strings, got: '" << grammar << "'";
-        } else {
-            KALDI_LOG << obj;
-
-            LanguageModelOptions opts;
-
-            opts.ngram_order = 2;
-            opts.discount = 0.5;
-
-            LanguageModelEstimator estimator(opts);
-            for (int i = 0; i < obj.length(); i++) {
-                bool ok;
-                string line = obj[i].ToString(ok);
-                if (!ok) {
-                    KALDI_ERR << "Expecting array of strings, got: '" << obj << "'";
-                }
-
-                std::vector<int32> sentence;
-                stringstream ss(line);
-                string token;
-                while (getline(ss, token, ' ')) {
-                    int32 id = model_->word_syms_->Find(token);
-                    if (id == kNoSymbol) {
-                        KALDI_WARN << "Ignoring word missing in vocabulary: '" << token << "'";
-                    } else {
-                        sentence.push_back(id);
-                    }
-                }
-                estimator.AddCounts(sentence);
-            }
-            g_fst_ = new StdVectorFst();
-            estimator.Estimate(g_fst_);
-
-            decode_fst_ = LookaheadComposeFst(*model_->hcl_fst_, *g_fst_, model_->disambig_);
-        }
+        UpdateGrammarFst(grammar);
     } else {
         KALDI_WARN << "Runtime graphs are not supported by this model";
     }
@@ -267,6 +229,96 @@ void Recognizer::SetSpkModel(SpkModel *spk_model)
     spk_feature_ = new OnlineMfcc(spk_model_->spkvector_mfcc_opts);
 }
 
+void Recognizer::SetGrm(char const *grammar)
+{
+    if (state_ == RECOGNIZER_RUNNING) {
+        KALDI_ERR << "Can't add speaker model to already running recognizer";
+        return;
+    }
+
+    if (!model_->hcl_fst_) {
+        KALDI_WARN << "Runtime graphs are not supported by this model";
+        return;
+    }
+
+    delete decode_fst_;
+
+    if (!strcmp(grammar, "[]")) {
+        decode_fst_ = LookaheadComposeFst(*model_->hcl_fst_, *model_->g_fst_, model_->disambig_);
+    } else {
+        UpdateGrammarFst(grammar);
+    }
+
+    samples_round_start_ += samples_processed_;
+    samples_processed_ = 0;
+    frame_offset_ = 0;
+
+    delete decoder_;
+    delete feature_pipeline_;
+    delete silence_weighting_;
+
+    silence_weighting_ = new kaldi::OnlineSilenceWeighting(*model_->trans_model_, model_->feature_info_.silence_weighting_config, 3);
+    feature_pipeline_ = new kaldi::OnlineNnet2FeaturePipeline (model_->feature_info_);
+    decoder_ = new kaldi::SingleUtteranceNnet3IncrementalDecoder(model_->nnet3_decoding_config_,
+            *model_->trans_model_,
+            *model_->decodable_info_,
+            *decode_fst_,
+            feature_pipeline_);
+
+    if (spk_model_) {
+        delete spk_feature_;
+        spk_feature_ = new OnlineMfcc(spk_model_->spkvector_mfcc_opts);
+    }
+
+    state_ = RECOGNIZER_INITIALIZED;
+}
+
+
+void Recognizer::UpdateGrammarFst(char const *grammar)
+{
+    json::JSON obj;
+    obj = json::JSON::Load(grammar);
+
+    if (obj.length() <= 0) {
+        KALDI_WARN << "Expecting array of strings, got: '" << grammar << "'";
+        return;
+    }
+
+    KALDI_LOG << obj;
+
+    LanguageModelOptions opts;
+
+    opts.ngram_order = 2;
+    opts.discount = 0.5;
+
+    LanguageModelEstimator estimator(opts);
+    for (int i = 0; i < obj.length(); i++) {
+        bool ok;
+        string line = obj[i].ToString(ok);
+        if (!ok) {
+            KALDI_ERR << "Expecting array of strings, got: '" << obj << "'";
+        }
+
+        std::vector<int32> sentence;
+        stringstream ss(line);
+        string token;
+        while (getline(ss, token, ' ')) {
+            int32 id = model_->word_syms_->Find(token);
+            if (id == kNoSymbol) {
+                KALDI_WARN << "Ignoring word missing in vocabulary: '" << token << "'";
+            } else {
+                sentence.push_back(id);
+            }
+        }
+        estimator.AddCounts(sentence);
+    }
+    g_fst_ = new StdVectorFst();
+    estimator.Estimate(g_fst_);
+
+    decode_fst_ = LookaheadComposeFst(*model_->hcl_fst_, *g_fst_, model_->disambig_);
+}
+
+
 bool Recognizer::AcceptWaveform(const char *data, int len)
 {
     Vector<BaseFloat> wave;
@@ -418,14 +470,23 @@ bool Recognizer::GetSpkVector(Vector<BaseFloat> &out_xvector, int *num_spk_frame
     return true;
 }
 
+// If we can't align, we still need to prepare for MBR
+static void CopyLatticeForMbr(CompactLattice &lat, CompactLattice *lat_out)
+{
+    *lat_out = lat;
+    RmEpsilon(lat_out, true);
+    fst::CreateSuperFinal(lat_out);
+    TopSortCompactLatticeIfNeeded(lat_out);
+}
 
 const char *Recognizer::MbrResult(CompactLattice &rlat)
 {
+
     CompactLattice aligned_lat;
     if (model_->winfo_) {
         WordAlignLattice(rlat, *model_->trans_model_, *model_->winfo_, 0, &aligned_lat);
     } else {
-        aligned_lat = rlat;
+        CopyLatticeForMbr(rlat, &aligned_lat);
     }
 
     MinimumBayesRisk mbr(aligned_lat);
@@ -739,7 +800,11 @@ const char* Recognizer::PartialResult()
         CompactLattice aligned_lat;
 
         clat = decoder_->GetLattice(decoder_->NumFramesInLattice(), false);
-        WordAlignLatticePartial(clat, *model_->trans_model_, *model_->winfo_, 0, &aligned_lat);
+        if (model_->winfo_) {
+            WordAlignLatticePartial(clat, *model_->trans_model_, *model_->winfo_, 0, &aligned_lat);
+        } else {
+            CopyLatticeForMbr(clat, &aligned_lat);
+        }
 
         MinimumBayesRisk mbr(aligned_lat);
         const vector<BaseFloat> &conf = mbr.GetOneBestConfidences();

src/recognizer.h

@@ -48,6 +48,7 @@ class Recognizer {
         ~Recognizer();
         void SetMaxAlternatives(int max_alternatives);
         void SetSpkModel(SpkModel *spk_model);
+        void SetGrm(char const *grammar);
         void SetWords(bool words);
         void SetPartialWords(bool partial_words);
         void SetNLSML(bool nlsml);
@@ -64,6 +65,7 @@ class Recognizer {
         void InitRescoring();
         void CleanUp();
         void UpdateSilenceWeights();
+        void UpdateGrammarFst(char const *grammar);
         bool AcceptWaveform(Vector<BaseFloat> &wdata);
         bool GetSpkVector(Vector<BaseFloat> &out_xvector, int *frames);
         const char *GetResult();

src/vosk_api.cc

@@ -121,6 +121,14 @@ void vosk_recognizer_set_spk_model(VoskRecognizer *recognizer, VoskSpkModel *spk
     ((Recognizer *)recognizer)->SetSpkModel((SpkModel *)spk_model);
 }
 
+void vosk_recognizer_set_grm(VoskRecognizer *recognizer, char const *grammar)
+{
+    if (recognizer == nullptr) {
+       return;
+    }
+    ((Recognizer *)recognizer)->SetGrm(grammar);
+}
+
 int vosk_recognizer_accept_waveform(VoskRecognizer *recognizer, const char *data, int length)
 {
     try {
@@ -195,10 +203,10 @@ void vosk_gpu_thread_init()
 #endif
 }
 
-VoskBatchModel *vosk_batch_model_new()
+VoskBatchModel *vosk_batch_model_new(const char *model_path)
 {
 #if HAVE_CUDA
-    return (VoskBatchModel *)(new BatchModel());
+    return (VoskBatchModel *)(new BatchModel(model_path));
 #else
     return NULL;
 #endif

src/vosk_api.h

@@ -146,6 +146,15 @@ VoskRecognizer *vosk_recognizer_new_grm(VoskModel *model, float sample_rate, con
 void vosk_recognizer_set_spk_model(VoskRecognizer *recognizer, VoskSpkModel *spk_model);
 
 
+/** Reconfigures recognizer to use grammar
+ *
+ * @param recognizer   Already running VoskRecognizer
+ * @param grammar      Set of phrases in JSON array of strings or "[]" to use default model graph.
+ *                     See also vosk_recognizer_new_grm
+ */
+void vosk_recognizer_set_grm(VoskRecognizer *recognizer, char const *grammar);
+
+
 /** Configures recognizer to output n-best results
  *
  * <pre>
@@ -243,7 +252,7 @@ int vosk_recognizer_accept_waveform_f(VoskRecognizer *recognizer, const float *d
  *  }
  * </pre>
  *
- * If alternatives enabled it returns result with alternatives, see also vosk_recognizer_set_alternatives().
+ * If alternatives enabled it returns result with alternatives, see also vosk_recognizer_set_max_alternatives().
  *
  * If word times enabled returns word time, see also vosk_recognizer_set_word_times().
  */
@@ -310,7 +319,7 @@ void vosk_gpu_thread_init();
 /** Creates the batch recognizer object
  *
  *  @returns model object or NULL if problem occured */
-VoskBatchModel *vosk_batch_model_new();
+VoskBatchModel *vosk_batch_model_new(const char *model_path);
 
 /** Releases batch model object */
 void vosk_batch_model_free(VoskBatchModel *model);

training/local/data_prep.sh

@@ -1 +0,0 @@
-../../../librispeech/s5/local/data_prep.sh

training/local/data_prep.sh

@@ -0,0 +1,85 @@
+#!/usr/bin/env bash
+
+# Copyright 2014  Vassil Panayotov
+#           2014  Johns Hopkins University (author: Daniel Povey)
+# Apache 2.0
+
+if [ "$#" -ne 2 ]; then
+  echo "Usage: $0 <src-dir> <dst-dir>"
+  echo "e.g.: $0 /export/a15/vpanayotov/data/LibriSpeech/dev-clean data/dev-clean"
+  exit 1
+fi
+
+src=$1
+dst=$2
+
+# all utterances are FLAC compressed
+if ! which flac >&/dev/null; then
+   echo "Please install 'flac' on ALL worker nodes!"
+   exit 1
+fi
+
+spk_file=$src/../SPEAKERS.TXT
+
+mkdir -p $dst || exit 1;
+
+[ ! -d $src ] && echo "$0: no such directory $src" && exit 1;
+[ ! -f $spk_file ] && echo "$0: expected file $spk_file to exist" && exit 1;
+
+
+wav_scp=$dst/wav.scp; [[ -f "$wav_scp" ]] && rm $wav_scp
+trans=$dst/text; [[ -f "$trans" ]] && rm $trans
+utt2spk=$dst/utt2spk; [[ -f "$utt2spk" ]] && rm $utt2spk
+spk2gender=$dst/spk2gender; [[ -f $spk2gender ]] && rm $spk2gender
+
+for reader_dir in $(find -L $src -mindepth 1 -maxdepth 1 -type d | sort); do
+  reader=$(basename $reader_dir)
+  if ! [ $reader -eq $reader ]; then  # not integer.
+    echo "$0: unexpected subdirectory name $reader"
+    exit 1;
+  fi
+
+  reader_gender=$(egrep "^$reader[ ]+\|" $spk_file | awk -F'|' '{gsub(/[ ]+/, ""); print tolower($2)}')
+  if [ "$reader_gender" != 'm' ] && [ "$reader_gender" != 'f' ]; then
+    echo "Unexpected gender: '$reader_gender'"
+    exit 1;
+  fi
+
+  for chapter_dir in $(find -L $reader_dir/ -mindepth 1 -maxdepth 1 -type d | sort); do
+    chapter=$(basename $chapter_dir)
+    if ! [ "$chapter" -eq "$chapter" ]; then
+      echo "$0: unexpected chapter-subdirectory name $chapter"
+      exit 1;
+    fi
+
+    find -L $chapter_dir/ -iname "*.flac" | sort | xargs -I% basename % .flac | \
+      awk -v "dir=$chapter_dir" '{printf "%s flac -c -d -s %s/%s.flac |\n", $0, dir, $0}' >>$wav_scp|| exit 1
+
+    chapter_trans=$chapter_dir/${reader}-${chapter}.trans.txt
+    [ ! -f  $chapter_trans ] && echo "$0: expected file $chapter_trans to exist" && exit 1
+    cat $chapter_trans >>$trans
+
+    # NOTE: For now we are using per-chapter utt2spk. That is each chapter is considered
+    #       to be a different speaker. This is done for simplicity and because we want
+    #       e.g. the CMVN to be calculated per-chapter
+    awk -v "reader=$reader" -v "chapter=$chapter" '{printf "%s %s-%s\n", $1, reader, chapter}' \
+      <$chapter_trans >>$utt2spk || exit 1
+
+    # reader -> gender map (again using per-chapter granularity)
+    echo "${reader}-${chapter} $reader_gender" >>$spk2gender
+  done
+done
+
+spk2utt=$dst/spk2utt
+utils/utt2spk_to_spk2utt.pl <$utt2spk >$spk2utt || exit 1
+
+ntrans=$(wc -l <$trans)
+nutt2spk=$(wc -l <$utt2spk)
+! [ "$ntrans" -eq "$nutt2spk" ] && \
+  echo "Inconsistent #transcripts($ntrans) and #utt2spk($nutt2spk)" && exit 1;
+
+utils/validate_data_dir.sh --no-feats $dst || exit 1;
+
+echo "$0: successfully prepared data in $dst"
+
+exit 0

training/run.sh

@@ -3,11 +3,13 @@
 . ./cmd.sh
 . ./path.sh
 
-stage=0
+stage=-1
+stop_stage=100
+
 . utils/parse_options.sh
 
 # Data preparation
-if [ $stage -le 0 ]; then
+if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
   data_url=www.openslr.org/resources/31
   lm_url=www.openslr.org/resources/11
   database=corpus
@@ -24,13 +26,13 @@ if [ $stage -le 0 ]; then
 fi
 
 # Dictionary formatting
-if [ $stage -le 1 ]; then
+if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
   local/prepare_dict.sh data/local/lm data/local/dict
   utils/prepare_lang.sh data/local/dict "<UNK>" data/local/lang data/lang
 fi
 
 # Extract MFCC features
-if [ $stage -le 2 ]; then
+if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
   for task in train; do
     steps/make_mfcc.sh --cmd "$train_cmd" --nj 10 data/$task exp/make_mfcc/$task $mfcc
     steps/compute_cmvn_stats.sh data/$task exp/make_mfcc/$task $mfcc
@@ -38,7 +40,7 @@ if [ $stage -le 2 ]; then
 fi
 
 # Train GMM models
-if [ $stage -le 3 ]; then
+if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
   steps/train_mono.sh --nj 10 --cmd "$train_cmd" \
     data/train data/lang exp/mono
 
@@ -65,12 +67,12 @@ if [ $stage -le 3 ]; then
 fi
 
 # Train TDNN model
-if [ $stage -le 4 ]; then
+if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then
   local/chain/run_tdnn.sh
 fi
 
 # Decode
-if [ $stage -le 5 ]; then
+if [ ${stage} -le 5 ] && [ ${stop_stage} -ge 5 ]; then
 
   utils/format_lm.sh data/lang data/local/lm/lm_tgsmall.arpa.gz data/local/dict/lexicon.txt data/lang_test
   utils/mkgraph.sh --self-loop-scale 1.0 data/lang_test exp/chain/tdnn exp/chain/tdnn/graph

travis/Dockerfile.dockcross

@@ -29,7 +29,7 @@ RUN cd /opt \
     && git clone -b v3.2.1  --single-branch https://github.com/alphacep/clapack \
     && echo ${OPENBLAS_ARGS} \
     && make -C OpenBLAS ONLY_CBLAS=1 ${OPENBLAS_ARGS} HOSTCC=gcc USE_LOCKING=1 USE_THREAD=0 all \
-    && make -C OpenBLAS PREFIX=$(pwd)/OpenBLAS/install install \
+    && make -C OpenBLAS ${OPENBLAS_ARGS} HOSTCC=gcc USE_LOCKING=1 USE_THREAD=0 PREFIX=$(pwd)/OpenBLAS/install install \
     && mkdir -p clapack/BUILD && cd clapack/BUILD && cmake .. \
     && make -j 10 -C F2CLIBS \
     && make -j 10 -C BLAS \

travis/Dockerfile.dockcross-manylinux

@@ -0,0 +1,39 @@
+ARG DOCKCROSS_IMAGE=dockcross/manylinux2014-aarch64
+FROM ${DOCKCROSS_IMAGE}
+
+LABEL description="A docker image for building portable Python linux binary wheels and Kaldi on other architectures"
+LABEL maintainer="contact@alphacephei.com"
+
+RUN yum -y install \
+    automake \
+    autoconf \
+    libtool \
+    libffi-devel \
+    && yum clean all
+
+ARG OPENBLAS_ARGS=
+RUN cd /opt \
+    && git clone -b vosk --single-branch https://github.com/alphacep/kaldi \
+    && cd kaldi/tools \
+    && git clone -b v0.3.20 --single-branch https://github.com/xianyi/OpenBLAS \
+    && git clone -b v3.2.1  --single-branch https://github.com/alphacep/clapack \
+    && echo ${OPENBLAS_ARGS} \
+    && make -C OpenBLAS ONLY_CBLAS=1 ${OPENBLAS_ARGS} HOSTCC=gcc USE_LOCKING=1 USE_THREAD=0 all \
+    && make -C OpenBLAS ${OPENBLAS_ARGS} HOSTCC=gcc USE_LOCKING=1 USE_THREAD=0 PREFIX=$(pwd)/OpenBLAS/install install \
+    && mkdir -p clapack/BUILD && cd clapack/BUILD && cmake .. \
+    && make -j 10 -C F2CLIBS \
+    && make -j 10 -C BLAS \
+    && make -j 10 -C SRC \
+    && find . -name "*.a" | xargs cp -t ../../OpenBLAS/install/lib \
+    && cd /opt/kaldi/tools \
+    && git clone --single-branch https://github.com/alphacep/openfst openfst \
+    && cd openfst \
+    && autoreconf -i \
+    && CFLAGS="-g -O3" ./configure --prefix=/opt/kaldi/tools/openfst --enable-static --enable-shared --enable-far --enable-ngram-fsts --enable-lookahead-fsts --with-pic --disable-bin --host=${CROSS_TRIPLE} --build=x86-linux-gnu \
+    && make -j 10 && make install \
+    && cd /opt/kaldi/src \
+    && sed -i "s:TARGET_ARCH=\"\`uname -m\`\":TARGET_ARCH=$(echo $CROSS_TRIPLE|cut -d - -f 1):g" configure \
+    && sed -i "s: -O1 : -O3 :g" makefiles/linux_openblas_arm.mk \
+    && ./configure --mathlib=OPENBLAS_CLAPACK --shared --use-cuda=no \
+    && make -j 10 online2 lm rnnlm \
+    && find /opt/kaldi -name "*.o" -exec rm {} \;

travis/Dockerfile.manylinux

@@ -9,6 +9,7 @@ RUN yum -y update && yum -y install \
     autoconf \
     libtool \
     cmake \
+    libffi-devel \
     && yum clean all
 
 RUN cd /opt \

travis/Dockerfile.win

@@ -1,4 +1,4 @@
-FROM debian:10.4
+FROM debian:11
 
 RUN apt-get update && \
     apt-get install -y --no-install-recommends \

travis/Dockerfile.win32

@@ -1,4 +1,4 @@
-FROM debian:10.4
+FROM debian:11
 
 RUN apt-get update && \
     apt-get install -y --no-install-recommends \

travis/build-dockcross-manylinux.sh

@@ -0,0 +1,7 @@
+#!/bin/bash
+
+set -e
+set -x
+
+docker build --build-arg="DOCKCROSS_IMAGE=dockcross/manylinux2014-aarch64" --build-arg="OPENBLAS_ARGS=TARGET=ARMV8" --file Dockerfile.dockcross-manylinux --tag alphacep/kaldi-dockcross-aarch64:latest .
+docker run --rm -v /home/shmyrev/travis/vosk-api/:/io alphacep/kaldi-dockcross-aarch64 /io/travis/build-wheels-dockcross.sh

travis/build-dockcross.sh

@@ -4,9 +4,13 @@ set -e
 set -x
 
 docker build --build-arg="DOCKCROSS_IMAGE=alphacep/dockcross-linux-armv7" --build-arg="OPENBLAS_ARGS=TARGET=ARMV7" --file Dockerfile.dockcross --tag alphacep/kaldi-dockcross-armv7:latest .
-docker build --build-arg="DOCKCROSS_IMAGE=dockcross/linux-arm64" --build-arg="OPENBLAS_ARGS=TARGET=ARMV8" --file Dockerfile.dockcross --tag alphacep/kaldi-dockcross-arm64:latest .
 docker build --build-arg="DOCKCROSS_IMAGE=dockcross/linux-x86" --build-arg="OPENBLAS_ARGS=TARGET=CORE2\ DYNAMIC_ARCH=1" --file Dockerfile.dockcross --tag alphacep/kaldi-dockcross-x86:latest .
+docker build --build-arg="DOCKCROSS_IMAGE=dockcross/linux-riscv64" --build-arg="OPENBLAS_ARGS=TARGET=RISCV64_GENERIC\ ARCH=riscv64" --file Dockerfile.dockcross --tag alphacep/kaldi-dockcross-riscv:latest .
 
 docker run --rm -v /home/shmyrev/travis/vosk-api/:/io alphacep/kaldi-dockcross-armv7 /io/travis/build-wheels-dockcross.sh
-docker run --rm -v /home/shmyrev/travis/vosk-api/:/io alphacep/kaldi-dockcross-arm64 /io/travis/build-wheels-dockcross.sh
 docker run --rm -v /home/shmyrev/travis/vosk-api/:/io alphacep/kaldi-dockcross-x86 /io/travis/build-wheels-dockcross.sh
+docker run --rm -v /home/shmyrev/travis/vosk-api/:/io alphacep/kaldi-dockcross-riscv /io/travis/build-wheels-dockcross.sh
+
+# We use manylinux (Centos-based image) for aarch64 instead
+# docker build --build-arg="DOCKCROSS_IMAGE=dockcross/linux-arm64" --build-arg="OPENBLAS_ARGS=TARGET=ARMV8" --file Dockerfile.dockcross --tag alphacep/kaldi-dockcross-arm64:latest .
+# docker run --rm -v /home/shmyrev/travis/vosk-api/:/io alphacep/kaldi-dockcross-arm64 /io/travis/build-wheels-dockcross.sh

travis/build-wheels-dockcross.sh

@@ -11,20 +11,27 @@ KALDI_ROOT=/opt/kaldi EXTRA_LDFLAGS="-latomic" make -j $(nproc)
 export VOSK_SOURCE=/opt/vosk-api
 case $CROSS_TRIPLE in
     *armv7-*)
-        export VOSK_ARCHITECTURE=armv7l
-        ;;
-    *aarch64-*)
-        export VOSK_ARCHITECTURE=aarch64
+        export VOSK_MACHINE=armv7l
+        export VOSK_ARCHITECTURE=32bit
         ;;
     *i686-*)
-        export VOSK_ARCHITECTURE=x86
+        export VOSK_MACHINE=x86
+        export VOSK_ARCHITECTURE=32bit
+        ;;
+    *aarch64-*)
+        export VOSK_MACHINE=aarch64
+        export VOSK_ARCHITECTURE=64bit
+        ;;
+    *riscv64-*)
+        export VOSK_MACHINE=riscv64
+        export VOSK_ARCHITECTURE=64bit
         ;;
 esac
 
 # Copy library to output folder
-mkdir -p /io/wheelhouse/vosk-linux-$VOSK_ARCHITECTURE
-cp /opt/vosk-api/src/*.so /opt/vosk-api/src/vosk_api.h /io/wheelhouse/vosk-linux-$VOSK_ARCHITECTURE
+mkdir -p /io/wheelhouse/vosk-linux-$VOSK_MACHINE
+cp /opt/vosk-api/src/*.so /opt/vosk-api/src/vosk_api.h /io/wheelhouse/vosk-linux-$VOSK_MACHINE
 
 # Build wheel
-python3 -m pip install requests tqdm
+python3 -m pip install requests tqdm srt websockets wheel
 python3 -m pip wheel /opt/vosk-api/python --no-deps -w /io/wheelhouse

travis/build-wheels-win.sh

@@ -18,6 +18,6 @@ cp /opt/vosk-api/src/*.{dll,lib} /opt/vosk-api/src/vosk_api.h /io/wheelhouse/vos
 
 # Build wheel and put to the output folder
 export VOSK_SOURCE=/opt/vosk-api
-export VOSK_PLATFORM=Windows
+export VOSK_SYSTEM=Windows
 export VOSK_ARCHITECTURE=64bit
 python3 -m pip -v wheel /opt/vosk-api/python --no-deps -w /io/wheelhouse

travis/build-wheels-win32.sh

@@ -18,6 +18,6 @@ cp /opt/vosk-api/src/*.{dll,lib} /opt/vosk-api/src/vosk_api.h /io/wheelhouse/vos
 
 # Build wheel and put to the output folder
 export VOSK_SOURCE=/opt/vosk-api
-export VOSK_PLATFORM=Windows
+export VOSK_SYSTEM=Windows
 export VOSK_ARCHITECTURE=32bit
 python3 -m pip -v wheel /opt/vosk-api/python --no-deps -w /io/wheelhouse

webjs/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vosk-js",
-  "version": "0.3.41",
+  "version": "0.3.45",
   "description": "Node binding for continuous voice recoginition through vosk-api.",
   "repository": {
     "type": "git",