Feature: Add SenseVoice as an ASR engine for faster subtitle generation

[LauraGPT]

Summary

Add SenseVoice as an alternative ASR engine for subtitle generation.

Why?

SmartSub generates subtitles from video — SenseVoice can make this faster and more accurate:

• 5x faster: Non-autoregressive model, generates all tokens in a single forward pass
• 50+ languages: Single 234MB model, no need to download separate models per language
• Better CJK accuracy: Lower CER than Whisper on Chinese, Japanese, Korean benchmarks
• No silence hallucination: Won't generate phantom text during quiet parts of video

Integration

from funasr import AutoModel

model = AutoModel(model="iic/SenseVoiceSmall", vad_model="fsmn-vad", device="cuda")
result = model.generate(input="audio.wav")
# Returns text with timestamps for each segment

For cross-platform (no Python): Sherpa-ONNX provides TypeScript/C++ bindings for SenseVoice.

References

• FunASR: https://github.com/modelscope/FunASR (16K+ stars)
• SenseVoice: https://github.com/FunAudioLLM/SenseVoice (8K+ stars)