> [!NOTE] > 本文档由 WeHub 基于上游 README 翻译整理,属于社区翻译,非官方中文文档。 > [English](./README.en.md) · [原始项目](https://github.com/MoonInTheRiver/DiffSinger) · [上游 README](https://github.com/MoonInTheRiver/DiffSinger/blob/HEAD/README.md) > 原作者、版权与许可证归属以原始项目及本仓库 LICENSE 文件为准。 # DiffSinger:基于浅层扩散机制的歌唱语音合成(Singing Voice Synthesis via Shallow Diffusion Mechanism) [![arXiv](https://img.shields.io/badge/arXiv-Paper-.svg)](https://arxiv.org/abs/2105.02446) [![GitHub Stars](https://img.shields.io/github/stars/MoonInTheRiver/DiffSinger?style=social)](https://github.com/MoonInTheRiver/DiffSinger) [![downloads](https://img.shields.io/github/downloads/MoonInTheRiver/DiffSinger/total.svg)](https://github.com/MoonInTheRiver/DiffSinger/releases) [![Hugging Face](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-blue?label=TTSDemo)](https://huggingface.co/spaces/NATSpeech/DiffSpeech) [![Hugging Face](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-blue?label=SVSDemo)](https://huggingface.co/spaces/Silentlin/DiffSinger) 本仓库是我们 AAAI-2022 [论文](https://arxiv.org/abs/2105.02446), 的官方 PyTorch 实现,其中我们提出了 DiffSinger(用于歌唱语音合成 Singing-Voice-Synthesis)和 DiffSpeech(用于文本转语音 Text-to-Speech)。 :tada: :tada: :tada: **更新**: - Sep.11, 2022::electric_plug: [DiffSinger-PN](docs/README-SVS-opencpop-pndm.md)。接入我们实验室 ICLR 2022 的插件 [PNDM](https://arxiv.org/abs/2202.09778),,可免费加速 DiffSinger。 - Jul.27, 2022:更新 [SVS](docs/README-SVS.md) 相关文档。新增简易推理 [A](docs/README-SVS-opencpop-cascade.md#4-inference-from-raw-inputs) 与 [B](docs/README-SVS-opencpop-e2e.md#4-inference-from-raw-inputs);新增运行于 [HuggingFace🤗 SVS](https://huggingface.co/spaces/Silentlin/DiffSinger). 的交互式 SVS。 - Mar.2, 2022:MIDI-B-version。 - Mar.1, 2022:用于歌声美化的 [NeuralSVB](https://github.com/MoonInTheRiver/NeuralSVB), 已发布。 - Feb.13, 2022:改进版代码框架 [NATSpeech](https://github.com/NATSpeech/NATSpeech), 已发布,其中包含 DiffSpeech 以及我们 NeurIPS-2021 工作 [PortaSpeech](https://openreview.net/forum?id=xmJsuh8xlq) 的实现。 - Jan.29, 2022:支持 MIDI-A-version SVS。 - Jan.13, 2022:支持 SVS,发布 PopCS 数据集。 - Dec.19, 2021:支持 TTS。[HuggingFace🤗 TTS](https://huggingface.co/spaces/NATSpeech/DiffSpeech) :rocket: **新闻**: - Feb.24, 2022:我们的新工作 NeuralSVB 被 ACL-2022 接收 [![arXiv](https://img.shields.io/badge/arXiv-Paper-.svg)](https://arxiv.org/abs/2202.13277). [演示页面](https://neuralsvb.github.io). - Dec.01, 2021:DiffSinger 被 AAAI-2022 接收。 - Sep.29, 2021:我们近期工作 `PortaSpeech: Portable and High-Quality Generative Text-to-Speech` 被 NeurIPS-2021 接收 [![arXiv](https://img.shields.io/badge/arXiv-Paper-.svg)](https://arxiv.org/abs/2109.15166) 。 - May.06, 2021:我们将 DiffSinger 提交至 Arxiv [![arXiv](https://img.shields.io/badge/arXiv-Paper-.svg)](https://arxiv.org/abs/2105.02446). ## 环境 1. 若你想使用 anaconda 环境: ```sh conda create -n your_env_name python=3.8 source activate your_env_name pip install -r requirements_2080.txt (GPU 2080Ti, CUDA 10.2) or pip install -r requirements_3090.txt (GPU 3090, CUDA 11.4) ``` 2. 或者,若你想使用 python 的 virtual env: ```sh ## 请先安装 Python 3.8。 python -m venv venv source venv/bin/activate # 安装依赖。 pip install -U pip pip install Cython numpy==1.19.1 pip install torch==1.9.0 pip install -r requirements.txt ``` ## 文档 - [运行 DiffSpeech(TTS 版本)](docs/README-TTS.md)。 - [运行 DiffSinger(SVS 版本)](docs/README-SVS.md)。 ## 概览 | Mel 流水线 | 数据集 | 音高输入 | F0 预测 | 加速方法 | 声码器 | | ------------------------------------------------------------------------------------------- | ---------------------------------------------------------| ----------------- | ------------- | --------------------------- | ----------------------------- | | [DiffSpeech (Text->F0, Text+F0->Mel, Mel->Wav)](docs/README-TTS.md) | [Ljspeech](https://keithito.com/LJ-Speech-Dataset/) | None | Explicit | Shallow Diffusion | HiFiGAN | | [DiffSinger (Lyric+F0->Mel, Mel->Wav)](docs/README-SVS-popcs.md) | [PopCS](https://github.com/MoonInTheRiver/DiffSinger) | Ground-Truth F0 | None | Shallow Diffusion | NSF-HiFiGAN | | [DiffSinger (Lyric+MIDI->F0, Lyric+F0->Mel, Mel->Wav)](docs/README-SVS-opencpop-cascade.md) | [OpenCpop](https://wenet.org.cn/opencpop/) | MIDI | Explicit | Shallow Diffusion | NSF-HiFiGAN | | [FFT-Singer (Lyric+MIDI->F0, Lyric+F0->Mel, Mel->Wav)](docs/README-SVS-opencpop-cascade.md) | [OpenCpop](https://wenet.org.cn/opencpop/) | MIDI | Explicit | Invalid | NSF-HiFiGAN | | [DiffSinger (Lyric+MIDI->Mel, Mel->Wav)](docs/README-SVS-opencpop-e2e.md) | [OpenCpop](https://wenet.org.cn/opencpop/) | MIDI | Implicit | None | Pitch-Extractor + NSF-HiFiGAN | | [DiffSinger+PNDM (Lyric+MIDI->Mel, Mel->Wav)](docs/README-SVS-opencpop-pndm.md) | [OpenCpop](https://wenet.org.cn/opencpop/) | MIDI | Implicit | PLMS | Pitch-Extractor + NSF-HiFiGAN | | [DiffSpeech+PNDM (Text->Mel, Mel->Wav)](docs/README-TTS-pndm.md) | [Ljspeech](https://keithito.com/LJ-Speech-Dataset/) | None | Implicit | PLMS | HiFiGAN | ## Tensorboard ```sh tensorboard --logdir_spec exp_name ```
Tensorboard
## 引用 @article{liu2021diffsinger, title={Diffsinger: Singing voice synthesis via shallow diffusion mechanism}, author={Liu, Jinglin and Li, Chengxi and Ren, Yi and Chen, Feiyang and Liu, Peng and Zhao, Zhou}, journal={arXiv preprint arXiv:2105.02446}, volume={2}, year={2021}} ## 致谢 * lucidrains 的 [denoising-diffusion-pytorch](https://github.com/lucidrains/denoising-diffusion-pytorch) * 官方 [PyTorch Lightning](https://github.com/PyTorchLightning/pytorch-lightning) * kan-bayashi 的 [ParallelWaveGAN](https://github.com/kan-bayashi/ParallelWaveGAN) * jik876 的 [HifiGAN](https://github.com/jik876/hifi-gan) * 官方 [espnet](https://github.com/espnet/espnet) * lmnt-com 的 [DiffWave](https://github.com/lmnt-com/diffwave) * keonlee9420 的 [Implementation](https://github.com/keonlee9420/DiffSinger). 特别感谢: * Openvpi 团队的维护:[DiffSinger](https://github.com/openvpi/DiffSinger). * 您的再创作与分享。