modelscope /FunASR
Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.
★ 21,033⑂ 2,129👁 113⑃ 2,001● 1,809▤ 87
pytorchspeech-recognitionparaformerpunctuationspeaker-diarizationvoice-activity-detectionasrmultilingual-asr
More from this author
+8
Daily
+109
Weekly
+329
Monthly
21,033
stars
2,129
forks
113
watchers
2,001
prs
1,809
issues
87
releases
README
(简体中文|English|日本語|한국어) Industrial speech recognition toolkit for offline, streaming, and edge deployment. ASR · VAD · punctuation · speaker pipelines · emotion and audio-event models · OpenAI-compatible serving Quick Start · Model selection · Models · Deployment matrix · Deployment hub · Docs · Benchmark · Contribute --- Quick Start Native Transformers For Fun-ASR-Nano transcription with the Hugging Face API, start with the Transformers 5.17.0 CPU quickstart. No FunASR toolkit or remote Python code is needed. Space · Notebook · Python / batch examples FunASR toolkit and pipelines No local setup? Open the Colab quickstart to transcribe a public sample or upload your own audio in a browser. Found FunASR useful? Star the project so more builders can find it. bash CPU-only installs can use the default PyPI wheels. pip install torch torchaudio pip install funasr For GPU quickstarts, install the PyTorch and torchaudio wheels that match your NVIDIA driver from pytorch.org before installing FunASR. After installation, confirm the GPU is visible: bash python - Deploy as API server: Local SenseVoice CPU recipe · Nano GPU serving and pinned vLLM setup Use with AI agents: MCP Server for Claude/Cursor · OpenAI API for LangChain/Dify/AutoGen Use with voice agents: OpenClaw realtime plugin (experimental; requires an unmerged upstream SDK patch) · Hugging Face speech-to-speech with SenseVoice (merged source checkout; not in v1.0.0) Why FunASR? FunASR is a toolkit: choose the task, checkpoint, and runtime separately. Support in one model or adapter does not imply support in every serving backend. | Task | Checkpoint or pipeline | Runtime entrypoint | Important limitation | |---|---|---|---| | File transcription with emotion/event tags | SenseVoiceSmall | Python AutoModel, CPU or GPU | Five-language checkpoint; tags do not identify speakers. | | LLM-based file transcription | Fun-ASR-Nano | AutoModel; split-engine AutoModelVLLM for the documented GPU path | Base Nano covers zh/en/ja and Chinese dialects/accents; timestamp support depends on checkpoint and path. | | Broader multilingual transcription | Fun-ASR-MLT-Nano | Python AutoModel | Separate 31-language checkpoint; do not transfer its coverage to base Nano. | | Chunked live transcription | Paraformer-zh-streaming | Streaming SDK or runtime WebSocket service | Use the streaming checkpoint and per-session cache, not an offline checkpoint. | | Speaker-aware file transcription | SenseVoiceSmall + FSMN-VAD + CAM++ | AutoModel with VAD and embedding clustering | Anonymous indices within a recording, not enrolled-speaker identification. | | Joint text, timestamps, and speakers | MOSS-Transcribe-Diarize, third-party OpenMOSS | FunASR adapter or upstream backend in the MOSS guide | Offline, recording-local anonymous labels; no external VAD/speaker pipeline for its unified path. | | Native CPU/edge transcription | Fun-ASR-Nano or SenseVoiceSmall GGU
Top issues
#3302增加Hotwords會造成語句變短及亂取代其他文字34 comments · 2026-07-22#3203建议给一个 roadmap28 comments · 2026-07-12#3479[Bug] Vulkan backend crashes on AMD GPUs on Windows (0xc0000005)22 comments · 2026-08-07#3038关于serve_realtime_ws.py服务的客户端测试达不到实时16 comments · 2026-06-29#3773中文断句的老问题,才发现代码调用和命令行方式运行的结果好像是有差别的,不一定适用所有音频文件哈11 comments · 2026-10-14#3825Fun-ASR-Nano-2512、SenseVoice复合流水线的 NPU 性能/兼容性9 comments · 2026-11-02#3496Fun-ASR-Nano Hugging Face checkpoint is missing 86 CTC tensors required for timestamps and diarization9 comments · 2026-08-14#3388SenseVoice-Small 增量微调:保留普通话+粤语的前提下新增潮汕话识别,如何避免灾难性遗忘?9 comments · 2026-07-23
Discussions
#3900License confirmation for bundling sherpa-onnx-paraformer-zh-small-2024-03-09 in a closed-source commercial Android app1 comments · 2026-11-14#3720FunASR v1.4.16: reliable mixed-device submodels and native Transformers guides0 comments · 2026-09-18#3642FunASR v1.4.14: portable source archives and verified runtimes0 comments · 2026-09-04#3586MOSS-Transcribe-Diarize joins the FunASR ecosystem / MOSS 一体化转写与说话人分段已接入 FunASR0 comments · 2026-08-30#3536FunASR 1.4.5: lighter Python install and nine-platform llama.cpp runtimes / 更轻安装与九平台运行时0 comments · 2026-08-27#3525Possible Paraformer timestamp issues in the Hugging Face full pipeline and GPU dynamic batching4 comments · 2026-08-25#3521FunASR 1.4.3: optional Silero VAD and fixed-K speaker clustering / 可选 Silero VAD 与固定 K 聚类0 comments · 2026-08-21#3513FunASR-nano-2512 + vLLM的AutoModelVLLM是否可以支持输出sentence_info句子级别的时间戳1 comments · 2026-08-17
Latest release
v1.4.26 v1.4.26 2026-11-12Final Publication Acceptance GitHub Release and PyPI 1.4.26 are published. Both official PyPI distributions were downloaded and match GitHub by SHA-256 and size: - funasr-1.4.26-py3-none-any.whl: 1015738 bytes; SHA-256 104e3704fed0ecfd6c08836468de147a2c6f05745281093b8a6dbe341d1ce831. - funasr-1.4.26.tar.gz: 1086172 bytes; SHA-256 763c7d860a49afd17e4b9c0636ff43f03f6519e68bd0a09a25d68318091e2cf8. FunASR 1.4.26 Source: cdd82b5cbfd05015f8fb2b167d61f25142c3a9e8 / tag v1.4.26. Fixes - 3874: response_format=text now returns the bare HTTP transcript instead of a JSON-quoted string. - 3872 / 3