Discussion: Integration with FunASR pipeline

[LauraGPT]

Discussion

FireRedVAD is an excellent VAD model! Since it supports FunASR integration, wanted to share some complementary use cases.

FunASR pipeline with FireRedVAD:

FunASR supports custom VAD models. FireRedVAD could be used as an alternative to the built-in FSMN-VAD for scenarios requiring higher accuracy:

from funasr import AutoModel

# FunASR with SenseVoice for ASR
model = AutoModel(
    model="iic/SenseVoiceSmall",
    vad_model="fsmn-vad",  # Could be replaced with FireRedVAD
)
result = model.generate(input="audio.wav")

Would love to see:

• Official integration example with FunASR's AutoModel pipeline
• Benchmark comparison: FireRedVAD vs FSMN-VAD on common datasets

Great work on the model! Looking forward to potential collaboration.

• FunASR: https://github.com/modelscope/FunASR (16K+ stars)