Feature Request: Add SenseVoice/FunASR as built-in ASR engine

[LauraGPT]

Feature Request

Duix-Avatar is an impressive open-source digital human toolkit. Suggesting SenseVoice / FunASR as a built-in ASR engine for voice interaction.

Why SenseVoice for digital humans?

• Non-autoregressive — fixed low latency (~100ms), critical for natural conversation with avatars
• 5x faster than Whisper — reduces total conversation latency significantly
• 234M params — lightweight, leaves GPU memory for avatar rendering
• Emotion detection — detects speaker emotions, enables responsive avatar expressions
• 50+ languages — single model for multilingual avatar interaction
• Built-in VAD — accurate endpoint detection for natural turn-taking

Digital human integration benefits

1. Lower latency = more natural conversation flow
2. Emotion awareness = avatar can react to user's emotional state
3. Audio event detection = avatar responds to laughter, applause
4. Streaming = real-time partial results for immediate avatar response

Quick integration

from funasr import AutoModel

model = AutoModel(model="iic/SenseVoiceSmall", vad_model="fsmn-vad")
result = model.generate(input=audio_chunk)
# text + emotion tags available immediately

• FunASR: https://github.com/modelscope/FunASR (16.6K stars)
• SenseVoice: https://github.com/FunAudioLLM/SenseVoice (8.3K stars)