FunASR

项目介绍

Industrial speech recognition. 170x faster than Whisper. 50+ languages.

No local setup? Open the Colab quickstart to transcribe a public sample or upload your own audio in a browser.

Flagship model — Fun-ASR-Nano (LLM-ASR, 31 languages; the default recommendation, needs a GPU):

On CPU (or for multilingual + emotion in one pass), use SenseVoice — which also returns speaker diarization and timestamps:

Output — structured text with speaker labels, timestamps, and punctuation:

That's it. One model, one call — VAD segmentation, speech recognition, punctuation, speaker diarization all happen automatically.

At scale, accelerate Fun-ASR-Nano with vLLM (batch processing):

Deploy as API…