Complementary: FunASR/SenseVoice — ASR for voice cloning pipelines

[LauraGPT]

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.

Summary

OpenVoice does voice cloning/synthesis. FunASR provides the complementary ASR front-end:

• 170x faster transcription — RTF 0.006-0.007
• 50+ languages (SenseVoice)
• Speaker diarization (CAM++) — useful for multi-speaker voice cloning
• OpenAI-compatible API — easy to chain ASR → voice cloning

In a voice pipeline: FunASR transcribes → OpenVoice clones voice → TTS generates. Both are Apache 2.0.

pip install funasr
funasr speaker_audio.wav --spk -f json  # Identify speakers
# → feed speaker audio to OpenVoice for cloning

GitHub: https://github.com/modelscope/FunASR (17.8K+ stars)