Belle-whisper-large-v3-zh-punct model for CTranslate2

This repository contains the conversion of the BELLE-2/Belle-whisper-large-v3-zh-punct model, which fine-tunes whisper-large-v3-zh to enhance Chinese punctuation mark capabilities. It demonstrates comparable performance to Belle-whisper-large-v3-zh on Chinese ASR benchmarks, including AISHELL1, AISHELL2, WENETSPEECH, and HKUST.

This model can be used in CTranslate2 or projects based on CTranslate2 such as faster-whisper.

Example

from faster_whisper import WhisperModel
model = WhisperModel("k1nto/Belle-whisper-large-v3-zh-punct-ct2")
segments, info = model.transcribe("audio.mp3")
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 BELLE-2/Belle-whisper-large-v3-zh-punct --output_dir Belle-whisper-large-v3-zh-punct-ct2 --copy_files tokenizer.json preprocessor_config.json --quantization float16