Qwen-Audio-3.1-ASR Technical Report
Speaker-labeled transcripts include timestamps and overlapping speech, with consistent speaker labels across turns.
Qwen-Audio-3.1-ASR retains support for 30 languages and 16 Chinese dialects. Its general audio capabilities cover speech and music captioning, sound event detection, temporal grounding, and audio question answering. The family includes variants for file transcription, low-latency streaming, interactive voice input, and non-streaming recognition. Speaker-diarized transcription and general audio understanding are available only in certain variants.
HF Daily Papers' note
Qwen-Audio-3.1-ASR retains support for 30 languages and 16 Chinese dialects. Its general audio capabilities cover speech and music captioning, sound event detection, temporal grounding, and audio question answering. The family includes variants for file transcription, low-latency streaming, interactive voice input, and non-streaming recognition. Speaker-diarized transcription and general audio understanding are available only in certain variants.
HF Daily Papers' note
score 6