xumo/whisper.cpp.custom

Model	Disk	Mem	SHA
whisper-medium-jp.jp.bin	1.5 GB	~2.6 GB	7658a6d66a1c56aad387c0b4c7472151d05fe0e6
Belle-whisper-large-v3-zh.zh.bin	2.9 GB	~4.7 GB	6e8ca64eae06da4be63d346730c731cb88413c69

# require openai/whisper and ggerganov/whisper.cpp
git clone https://github.com/openai/whisper
git clone https://github.com/ggerganov/whisper.cpp

# convert models on huggingface

git clone https://huggingface.co/Jingmiao/whisper-small-chinese_base
python whisper.cpp/models/convert-h5-to-ggml.py ./whisper-small-chinese_base ./whisper .
mv ggml-model.bin whisper-small-chinese_base.zh.bin

git clone https://huggingface.co/BELLE-2/Belle-whisper-large-v2-zh
python whisper.cpp/models/convert-h5-to-ggml.py ./Belle-whisper-large-v2-zh ./whisper .
mv ggml-model.bin Belle-whisper-large-v2-zh.zh.bin

# recommend
git clone https://huggingface.co/BELLE-2/Belle-whisper-large-v3-zh
python whisper.cpp/models/convert-h5-to-ggml.py ./Belle-whisper-large-v3-zh ./whisper .
mv ggml-model.bin Belle-whisper-large-v3-zh.zh.bin

git clone https://huggingface.co/vumichien/whisper-medium-jp
python whisper.cpp/models/convert-h5-to-ggml.py ./whisper-medium-jp ./whisper .
mv ggml-model.bin whisper-medium-jp.jp.bin


cd whisper.cpp
make quantize
./quantize whisper-medium-jp.jp.bin whisper-medium-jp.jp.bin q5_0