Qwen-ASR

0.0.6 verified Mon Apr 27 auth: no python

Qwen-ASR is a Python package for automatic speech recognition (ASR) using the Qwen3 model family. Version 0.0.6 supports real-time and file-based transcription with speaker diarization. Release cadence is irregular; in early development.

pip install qwen-asr

Common errors

error ModuleNotFoundError: No module named 'qwen_asr' ↓

cause Installed wrong package name or version mismatch.

fix

Run pip install qwen-asr==0.0.6. Verify with pip list | grep qwen-asr.

error AttributeError: module 'qwen_asr' has no attribute 'QwenASR' ↓

cause Using an outdated import (e.g., `from qwen_asr import ASR`).

fix

Use from qwen_asr import QwenASR (capital Q, capital A).

error ValueError: The api_key must be provided ↓

cause Missing API key for cloud-based inference.

fix

Set environment variable QWEN_API_KEY or pass the key directly: QwenASR(api_key='your_key').

Warnings

gotcha The QwenASR constructor requires an `api_key` argument; omitting it or passing a wrong value leads to silent fallback to a local model if available, otherwise fails. ↓

fix Always set the environment variable QWEN_API_KEY or pass api_key explicitly.

breaking Version 0.0.6 changed the import path from `qwen_asr` (previously it may have been `qwenasr`). ↓

fix Use `from qwen_asr import ...` instead of `import qwenasr`.

deprecated The `transcribe()` method's `return_dict` argument is deprecated; output structure changed in 0.0.6. ↓

fix Remove `return_dict` parameter; use plain call: `model.transcribe('file.wav')`.

Imports

QwenASR

wrong

from qwen_asr import ASR

correct

from qwen_asr import QwenASR

Class name is QwenASR, not ASR

AudioInput
```
from qwen_asr import AudioInput
```

Quickstart

Initialize QwenASR with an API key and transcribe an audio file.

from qwen_asr import QwenASR
import os

api_key = os.environ.get('QWEN_API_KEY', '')
model = QwenASR(api_key=api_key)
transcription = model.transcribe('audio.wav')
print(transcription)