Feature request: Add FunASR for voice input transcription

[LauraGPT]

Feature Request

Add FunASR/SenseVoice as a self-hosted STT option for voice input in Chat UI.

Why

Chat UI supports voice input via browser speech recognition. A server-side STT option using FunASR would provide:

• Better accuracy: SenseVoice outperforms browser speech recognition, especially for non-English languages
• Self-hosted: Aligns with Chat UI's self-hosted philosophy
• 50+ languages with emotion detection
• OpenAI-compatible API: funasr-server serves POST /v1/audio/transcriptions

Integration

# Start FunASR server alongside Chat UI
funasr-server --device cuda --port 8000

References

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