TOOLORA / Local reading edition
FunASR
Compose ASR, VAD, punctuation, timestamps and speaker clustering into offline, streaming or edge speech-processing services.
Overview
FunASR is a speech-processing toolkit whose AutoModel interface loads recognition, VAD, punctuation, emotion or speaker models into offline and streaming workflows. Pipelines produce text, timestamps and anonymous speaker indices within a recording; language coverage and hardware requirements depend on the selected models. SenseVoiceSmall with FSMN-VAD and CAM++ forms a diarization pipeline rather than a single model’s native capability. Toolkit code uses MIT, while model terms are separate.
Official sources and references
- Official websitehttps://github.com/modelscope/FunASR
- Documentationhttps://www.funasr.com/
Descriptions summarize public information; consult official websites for features, pricing and regional availability.