Qwen-Audio-3.1-ASR adds diarization and broader audio understanding
Original title: Qwen-Audio-3.1-ASR Technical Report
The family keeps 30 languages and 16 Chinese dialects, with file, streaming, and interactive variants.
The Qwen-Audio-3.1-ASR report presents a family that extends Qwen-Audio-3.0-ASR with stronger speech recognition, speaker-diarized transcription, and general audio understanding. Diarized output includes speaker labels, timestamps, and text, including overlapping speech. Variants cover file transcription, low-latency streaming, interactive voice input, and non-streaming recognition; diarization and audio understanding appear only in some variants.
Unconfirmed: Capability claims come from the authors' technical report, without independent confirmation.
- Controls include history conditioning, hotwords, and transcription polishing.
- Audio tasks cover captioning, grounding, event detection, and question answering.
- Evaluations span conventional, multilingual, dialectal, and diarized speech.
- Diarization and general audio understanding are limited to certain variants.
- AI models
- Qwen-Audio-3.1-ASRQwen-Audio-3.0-ASR
From the original article
1 min read at huggingface.co
We present Qwen-Audio-3. 1-ASR, a family of audio-language models that builds upon its predecesor Qwen-Audio-3. 0-ASR with improved speech recognition and expanded capabilities in speaker-diarized transcription and general audio understanding.
By Chuanmeng Bian, Daren Chen, Peixin Chen · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.