Audio
PaddleSpeech
Easy-to-use Speech Toolkit including Self-Supervised Learning model, SOTA/Streaming ASR with punctuation, Streaming TTS with text…
★ 12.7k starsApache-2.0
View guide →A complete open-source speech recognition toolkit - ASR, VAD, diarization and serving
FunASR is an open-source audio project by modelscope (modelscope/FunASR on GitHub). It has 20,613 GitHub stars and 2,066 forks, is released under the MIT license, and was last updated on GitHub on 8 October 2026.
Decide which tasks you need - transcription alone, or VAD, punctuation, speakers and emotion as well.
Install the funasr package from PyPI after installing PyTorch for your hardware.
Install torch/torchaudio, then funasr, and let it download the model checkpoints you choose.
Run the AutoModel pipeline or the CLI on a sample recording and inspect the segments and timestamps.
Combine ASR with FSMN-VAD segmentation, punctuation restoration and CAM++ speaker diarization in one AutoModel call.
Run real-time streaming recognition chunk by chunk, or high-accuracy offline transcription.
Includes Paraformer, SenseVoice and the Fun-ASR Nano family, with a 31-language multilingual checkpoint.
SenseVoice models add emotion recognition and audio-event tags alongside the transcript.
Ships an MCP server for Claude/Cursor and an OpenAI-compatible API for LangChain, Dify and AutoGen style integrations.
Transcribe from the command line with JSON output, SRT subtitles, speaker labels and timestamps.
Get the source code from GitHub and enter the project directory.
git clone https://github.com/modelscope/FunASR.git
cd FunASRInstall PyTorch first (CPU wheels are fine to start), then the toolkit itself.
pip install torch torchaudio
pip install funasrThe README's production example combines VAD, ASR, punctuation and speaker models in one pipeline.
from funasr import AutoModel
model = AutoModel(model="paraformer-zh", vad_model="fsmn-vad", punc_model="ct-punc", spk_model="cam++")
# res = model.generate(input="audio.wav") - see README usageIt is maintained by the ModelScope community (Alibaba) and contributors on GitHub.
The toolkit is MIT licensed and free; pretrained model weights are licensed separately, so check the license on each model card you download.
No API key is needed - models run locally on your CPU or GPU. Costs are only your own compute.
Compared with OpenAI Whisper, FunASR offers a broader production pipeline (VAD, punctuation, diarization, emotion) and strong Chinese/multilingual models, with serving options aimed at deployment.
Install torch and funasr with pip, load an AutoModel pipeline from the README, and transcribe a sample WAV file.
Easy-to-use Speech Toolkit including Self-Supervised Learning model, SOTA/Streaming ASR with punctuation, Streaming TTS with text…
Speech-to-text, text-to-speech, speaker diarization, speech enhancement, source separation, and VAD using next-gen Kaldi with onnxruntime without Internet co…
Automatic Speech Recognition with Speaker Diarization based on OpenAI Whisper
Deep learning toolkit for text-to-speech, battle-tested in research and production
A PyTorch-based Speech Toolkit
Obs-localvocal — obs plugin for local speech recognition and captioning, open-source.
The most-starred other Audio projects in the LibreAI catalogue — the realistic shortlist if FunASR is not the right fit.
利用 AI 大模型和自动化工作流，根据主题或关键词一键生成高清短视频。Generate HD short videos from a topic or keyword with an automated AI workflow.
Robust Speech Recognition via Large-Scale Weak Supervision
1 min voice data can also be used to train a good TTS model! (few shot voice cloning)
Clone a voice from 5 seconds of audio and generate arbitrary speech in real time
Port of OpenAI's Whisper model in C/C++
Free and open — grab the source, star the repo, and start creating.