# Qwen3-ASR

Qwen3-ASR is an open-source series of ASR models developed by the Qwen team at Alibaba Cloud, supporting stable multilingual speech/music/song recognition, language detection and timestamp prediction.

Repository: https://github.com/QwenLM/Qwen3-ASR
Canonical: https://ross.abutalabs.com/products/qwen3-asr
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-06-26T08:20:17+00:00

## Health v2 (maintenance only)
Score: 55/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 89, release rhythm 35, longevity 15
- inputs: {"age_days": 217, "days_push": 68, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3423, forks 349 (observed 2026-08-28T04:08:03.828584+00:00)

## What it is
Qwen3-ASR is a family of open-source speech recognition models from Alibaba's Qwen team, supporting ASR and language identification across 52 languages plus a forced-alignment model for text-speech alignment in 11 languages. It ships as a Python package with Transformers and vLLM backends, streaming inference, fine-tuning support, and a companion toolkit for high-throughput API transcription of long audio.

## Use cases
- transcribe audio to text in multiple languages
- add subtitles with timestamps to videos
- detect the spoken language of an audio clip
- transcribe songs and music vocals
- align transcripts with audio using forced alignment
- batch transcribe long audio files
- fine-tune a speech recognition model on custom data

## When to choose
- you need multilingual speech recognition with language detection and timestamps
- you want open-weight ASR models you can self-host with vLLM or Transformers
- you need forced alignment between text and speech
- you transcribe music or songs, not just speech

## When to avoid
- you need a tiny on-device model for mobile or embedded hardware
- you only need simple English-only transcription with minimal setup
- you lack GPU resources and cannot use a hosted API

## Facets
- artifact type: library
- maturity: active
- function: speech-recognition, machine-learning, llm-inference
- domain: speech-processing, machine-learning
- platform: python, cross-platform
- tags: asr, forced-alignment, multilingual, timestamps, language-identification, vllm, transformers, music-recognition, fine-tuning, natural-language-processing, audio, gpu, docker

## Member repositories
- QwenLM/Qwen3-ASR (main) score 55
- QwenLM/Qwen3-ASR-Toolkit (sdk) score 52

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:03.828584+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:38:05.361600+00:00, confidence not recorded.
  - readme: https://github.com/QwenLM/Qwen3-ASR (fetched 2026-08-28T04:08:03.828584+00:00, sha 9783bc33e2e8)
  - registry_pypi: https://pypi.org/pypi/qwen3-asr-toolkit/json (fetched 2026-08-29T09:32:35.887398+00:00, sha ab2e61983e5d)
- Data as of 2026-08-30T08:39:29.467469+00:00.