RoachLin/kotoba-whisper-v2.2-faster

Whisper kotoba-whisper-v2.2 model for CTranslate2

This repository contains the conversion of kotoba-tech/kotoba-whisper-v2.2 to the CTranslate2 model format.
This model can be used in CTranslate2 or projects based on CTranslate2 such as faster-whisper.

Example

Install library and get a sample audio.

pip install faster-whisper

Inference with the kotoba-whisper-v2.2-faster.

from faster_whisper import WhisperModel

if __name__ == '__main__':
    model = WhisperModel(model_size_or_path="./kotoba-whisper-v2.2-faster", device="cuda", compute_type="float32", local_files_only=True)
    segments, info = model.transcribe(audio="./123.wav", language="ja", chunk_length=5, condition_on_previous_text=False, hotwords="ノイミー")

    for segment in segments:
        print("[%.2fs -> %.2fs] %s" % (segment.start, segment.end, segment.text))

Conversion details

The original model was converted with the following command:

ct2-transformers-converter --model kotoba-tech/kotoba-whisper-v2.2 --output_dir kotoba-whisper-v2.2-faster --copy_files tokenizer.json preprocessor_config.json --quantization float32

Note that the model weights are saved in FP32. This type can be changed when the model is loaded using the compute_type option in CTranslate2.

More information

For more information about the kotoba-whisper-v2.0, refer to the original model card.