DocumentationQwen3-ASR Languages and Dialects
Install the local on-device Qwen3-ASR model and review its full list of 30 supported global languages and 22 Chinese dialects.
Kapinote provides out-of-the-box support for the open-source Qwen3-ASR 0.6B local on-device transcription model. Once downloaded to your Mac, speech recognition runs completely offline without consuming cloud quotas or transmitting audio data externally.
Setup steps
- Open Settings → Speech.
- Select Qwen3-ASR 0.6B from the model dropdown.
- Start the one-click model download (~1.2 GB) and keep Kapinote open until the download and checksum verify.
- Select your preferred language (or keep automatic detection) and run the built-in test.
- Record a short test sample containing accents and technical vocabulary common in your meetings.
Note: Initial setup requires an internet connection and at least 2 GB of free disk space. Real-time transcription speed depends on your Mac's Apple silicon configuration (M-series unified memory and GPU cores).
30 officially supported languages
According to the official Qwen3-ASR repository specifications, the model is trained and optimized for the following 30 languages:
| Language & Code | Language & Code | Language & Code | Language & Code |
|---|---|---|---|
Chinese (zh) | English (en) | Cantonese (yue) | Arabic (ar) |
German (de) | French (fr) | Spanish (es) | Portuguese (pt) |
Indonesian (id) | Italian (it) | Korean (ko) | Russian (ru) |
Thai (th) | Vietnamese (vi) | Japanese (ja) | Turkish (tr) |
Hindi (hi) | Malay (ms) | Dutch (nl) | Swedish (sv) |
Danish (da) | Finnish (fi) | Polish (pl) | Czech (cs) |
Filipino (fil) | Persian (fa) | Greek (el) | Hungarian (hu) |
Macedonian (mk) | Romanian (ro) |
22 officially supported Chinese dialects & accents
Qwen3-ASR offers exceptional adaptability for Chinese regional dialects and accents:
- Northern & Central Plain: Dongbei, Hebei, Tianjin, Shandong, Henan, Shaanxi, Shanxi
- Southwestern & Central: Sichuan, Hubei, Hunan, Guizhou, Yunnan, Jiangxi, Anhui
- Northwestern: Gansu, Ningxia
- Southern & Southeastern: Cantonese (Hong Kong accent), Cantonese (Guangdong accent), Wu, Minnan, Fujian, Zhejiang
Language support indicates the acoustic model is specifically trained on these accents. Real-world accuracy still depends on microphone quality, background noise, multi-speaker crosstalk, and vocabulary complexity. See the official Qwen3-ASR GitHub repository for ongoing updates.
Tips for better local transcription
- Microphone Placement: Use a dedicated headset or keep your Mac's microphone close to speakers to minimize room reverberation.
- Language Hints: If a meeting is primarily conducted in one language, explicitly select it rather than relying on automatic language detection.
- Custom Vocabulary: Add specialized project codenames, team member names, and acronyms in Custom vocabulary. While local Qwen transcription runs acoustically on-device, this vocabulary is supplied to downstream AI summary generation to guarantee perfect spelling in notes.