Paraformer Chinese ASR

Fast Mandarin and Chinese–English speech recognition with funasr/paraformer-zh.

Audio

Optional hotwords

Separate names or domain terms with spaces.

Enable speaker diarization

Uses CAM++ to label speakers; adds processing time.

Transcript

Word-level timestamps

Diarized segments

Download JSON

Timestamps are derived from Paraformer's native CIF alignment. Times are also retained as integer milliseconds in the JSON export.