Feature request: Add FunASR/SenseVoice as alternative ASR backend

[LauraGPT]

Note

License and capability clarification (2026-07-14): FunASR is a toolkit, not a single checkpoint. The FunASR and SenseVoice repository source code is MIT; model weights follow each model card. SenseVoiceSmall supports Chinese, Cantonese, English, Japanese, and Korean, and its weights use the linked FunASR Model Open Source License Agreement. Fun-ASR-Nano-2512 is Apache-2.0. Language coverage, punctuation, and performance depend on the selected model and runtime configuration.

Feature Request

Add FunASR/SenseVoice as an alternative to Whisper for transcription.

Why

• SenseVoice (234M params): 50+ languages, 5x faster than Whisper-small, non-autoregressive (no hallucination)
• Built-in speaker diarization (cam++): No need for separate diarization pipeline
• Complete pipeline in one package: VAD + ASR + punctuation + timestamps + speaker diarization
• OpenAI-compatible API: funasr-server serves POST /v1/audio/transcriptions — drop-in replacement

Quick start

pip install funasr
from funasr import AutoModel
model = AutoModel(model="iic/SenseVoiceSmall")
result = model.generate(input="audio.wav")

References

• FunASR — 16K+ stars
• SenseVoice — 8K+ stars