CUHK Cantonese ASR
Chinese University of Hong Kong — Cantonese speech research and open checkpoints.
Chinese University of Hong Kong — Cantonese speech research and open checkpoints.
Best for cantonese transcription for Hong Kong and Guangdong-area journalism and broadcast. Pricing: free.
What it is
CUHK's speech group has published Cantonese-tuned ASR checkpoints that handle the language's tonal phonology and code-switching with Mandarin and English more reliably than stock multilingual Whisper. Cantonese remains weakly supported by big-cloud STT, so these releases are the practical baseline for Hong Kong and Guangdong-area transcription projects. Best fit when the buyer is cantonese transcription for hong kong and guangdong-area journalism and broadcast. The honest caveat: research-grade releases; production hardening is integrator's responsibility. As with any open-weights release, the integrator owns hosting, scaling, and SLA — but the licensing cost is zero and the model can be fine-tuned on in-house audio.
Watch out for: Research-grade releases; production hardening is integrator's responsibility.
Install / use
huggingface.co search 'cantonese asr' for model cardsFeatures
| Speaker diarization | No |
| Word-level timestamps | Yes |
| Streaming / real-time | No |
| Languages supported | 1 |
| HIPAA eligible | No |
CUHK Cantonese ASR vs Whipscribe
| Feature | CUHK Cantonese ASR | Whipscribe |
|---|---|---|
| Category | Open source | Transcription APIs |
| Pricing | free | free beta |
| Speaker diarization | No | Yes |
| Word timestamps | Yes | Yes |
| Streaming | No | No |
| Languages | 1 | 99 |
| Platforms | Linux | Web, API, MCP |
Alternatives to CUHK Cantonese ASR
Whipscribe is a managed faster-whisper + whisperX service. If you want transcripts without running infrastructure, paste a URL or drop a file in the form below — you'll have a transcript in seconds.