FunASR v1.3.29: SenseVoice VAD timestamps and complete runtime downloads

[LauraGPT]

FunASR v1.3.29 is the current weekly release for SenseVoice segment timestamps, reliable realtime transcription, and self-contained llama.cpp deployment.

Fast path

python -m pip install -U "funasr==1.3.29"

What changed

• sentence_timestamp=True now returns each SenseVoice VAD region through sentence_info when token timestamps and a punctuation model are unavailable.
• Subtitle, clipping, and GUI clients receive usable start / end boundaries instead of an empty timeline.
• The release retains the v1.3.28 realtime fixes: stable final text after VAD-locked regressions and short-tail decoding on STOP.
• A real public-package CPU run with iic/SenseVoiceSmall returned the expected transcript and a 610-5530 ms segment.

The focused VAD, subtitle, CLI, and OpenAI-compatible API validation passed 30 tests before release.

Downloads and demos

• PyPI wheel and source distribution: https://pypi.org/project/funasr/1.3.29/
• Signed release plus nine Linux, macOS, and Windows llama.cpp packages: https://github.com/modelscope/FunASR/releases/tag/v1.3.29
• Fun-ASR-Nano runtime v0.1.9 with the same nine platform packages: https://github.com/QwenAudio/Fun-ASR/releases/tag/runtime-llamacpp-v0.1.9
• SenseVoice runtime v0.1.9 with the same nine platform packages and a post-security-fix source snapshot: https://github.com/QwenAudio/SenseVoice/releases/tag/runtime-llamacpp-v0.1.9
• Public FunClip demo running funasr==1.3.29: https://huggingface.co/spaces/FunAudioLLM/FunClip
• FunClip v2.1.0 source release (tar.gz, zip, and SHA256SUMS): https://github.com/modelscope/FunClip/releases/tag/v2.1.0
• FunClip v2.1.0 technical guide: https://www.funasr.com/en/blog/funclip-v2-1-0-video-clipping-release.html
• FunClip v2.1.0 中文技术文章: https://www.funasr.com/blog/funclip-v2-1-0-video-clipping-release.html
• Realtime deployment and subtitle guide: https://www.funasr.com/en/blog/funasr-v1-3-28-realtime-websocket-subtitles.html
• 中文实时服务与字幕说明: https://www.funasr.com/blog/funasr-v1-3-28-realtime-websocket-subtitles.html
• Official Dify Marketplace plugin: https://marketplace.dify.ai/plugin/langgenius/funasr

Related repositories:

• FunASR: https://github.com/modelscope/FunASR
• Fun-ASR-Nano: https://github.com/QwenAudio/Fun-ASR
• SenseVoice: https://github.com/QwenAudio/SenseVoice
• FunClip: https://github.com/modelscope/FunClip

If you test this release in a real application, please share your hardware, selected model, latency, and deployment friction here. Concrete reports help us turn support questions into fixes and examples.