Feature Request: Add FunASR as ASR parser option

[!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.

Hi! Leon is a great personal assistant with a pluggable ASR architecture.

I'd like to suggest adding FunASR (16K+ stars) as an ASR parser alongside the existing Coqui, Google Cloud, and Watson options:

Why FunASR for Leon:

• 170x real-time GPU speed — near-instant response for voice commands
• Streaming ASR: Paraformer-streaming model designed for real-time interaction with low latency
• 50+ languages: SenseVoice model with automatic language detection
• Built-in VAD + punctuation: No separate preprocessing needed
• Fully local/offline: No API keys or cloud dependency, privacy-first
• OpenAI-compatible API: funasr-server --device cuda serves at /v1/audio/transcriptions

FunASR follows the same local-first philosophy as Leon and would complement the existing parsers well — especially for users who need fast, accurate, multilingual recognition without cloud services.

from funasr import AutoModel
model = AutoModel(model="iic/SenseVoiceSmall")
result = model.generate(input="audio.wav")
print(result[0]["text"])

Happy to help implement the parser!