whisper-tiny-cantonese-ggml

whisper-tiny-cantonese-ggml

GGML conversion of Oblivion208/whisper-tiny-cantonese for use with whisper.cpp

Original Model

This is a GGML-converted version of Oblivion208/whisper-tiny-cantonese.

Original model credit: Oblivion208/whisper-tiny-cantonese

All credit for the fine-tuning work goes to the original authors. This repository only provides the GGML conversion for use with whisper.cpp and mobile applications.

Model Details

• Base Model: OpenAI Whisper Base
• Format: GGML (whisper.cpp compatible)
• Language: yue
• File Size: 74.09 MB
• Model File: ggml-tiny-yue.bin
• Conversion: Converted using whisper.cpp's convert-h5-to-ggml.py

Usage

With whisper.cpp

# Clone whisper.cpp
git clone https://github.com/ggerganov/whisper.cpp
cd whisper.cpp

# Build
make

# Download this model
wget https://huggingface.co/wabisabisocial/whisper-tiny-cantonese-ggml/resolve/main/ggml-tiny-yue.bin

# Run transcription
./main -m ggml-tiny-yue.bin -f audio.wav -l yue

With React Native (whisper.rn)

import WhisperContext from 'whisper.rn';

const context = await WhisperContext.initWithUrl(
  'https://huggingface.co/wabisabisocial/whisper-tiny-cantonese-ggml/resolve/main/ggml-tiny-yue.bin'
);

const { result } = await context.transcribe(audioPath, {
  language: 'yue',
});

Direct Download

wget https://huggingface.co/wabisabisocial/whisper-tiny-cantonese-ggml/resolve/main/ggml-tiny-yue.bin

Performance

For performance metrics and evaluation results, please refer to the original model.

License

This model is released under the Apache 2.0 License, matching the original model's license.

Per Apache 2.0 requirements:

• ✅ Commercial use allowed
• ✅ Modification allowed
• ✅ Distribution allowed
• ⚠️ Must include attribution to original authors
• ⚠️ Must include copy of license

Citation

If you use this model, please cite both the original model and OpenAI Whisper:

@misc{whisper-yue-ggml-wabisabisocial,
  author = {wabisabisocial},
  title = {whisper-tiny-cantonese-ggml},
  year = {2026},
  publisher = {Hugging Face},
  note = {GGML conversion of Oblivion208/whisper-tiny-cantonese},
  howpublished = {\url{https://huggingface.co/wabisabisocial/whisper-tiny-cantonese-ggml}}
}

@article{radford2022whisper,
  title = {Robust Speech Recognition via Large-Scale Weak Supervision},
  author = {Radford, Alec and Kim, Jong Wook and Xu, Tao and Brockman, Greg and McLeavey, Christine and Sutskever, Ilya},
  journal = {arXiv preprint arXiv:2212.04356},
  year = {2022}
}

Acknowledgments

• Original Whisper model: OpenAI Whisper
• GGML conversion tools: whisper.cpp
• Fine-tuned model: Oblivion208/whisper-tiny-cantonese