Breeze-ASR-25 int8 (CTranslate2)

這是 MediaTek-Research/Breeze-ASR-25 的 CTranslate2 int8 版本,可由 faster-whisper 載入。模型沒有額外微調。

Breeze-ASR-25 基於 Whisper-large-v2,主要針對台灣華語、繁體中文語境、華英句內/句外混用與字幕時間對齊進行強化。

使用方式

python -m pip install faster-whisper==1.2.1
from faster_whisper import WhisperModel

model = WhisperModel(
    "YOUR_NAMESPACE/Breeze-ASR-25-int8-CT2",
    device="cpu",
    compute_type="int8",
)

segments, info = model.transcribe(
    "audio.wav",
    language="zh",
    beam_size=5,
)

print(f"language={info.language} probability={info.language_probability:.3f}")
for segment in segments:
    print(f"[{segment.start:.2f}s -> {segment.end:.2f}s] {segment.text}")

使用 CUDA 時可將 device 改成 cuda;支援 FP16 的 GPU 可搭配 compute_type="int8_float16"

模型資訊

這是 CTranslate2 runtime artifact,不能用 transformers.AutoModel* 直接載入。

授權與引用

來源模型以 Apache License 2.0 釋出。使用前請一併閱讀來源模型卡

@article{chou2025selfrefiningframeworkenhancingasr,
  title={A Self-Refining Framework for Enhancing ASR Using TTS-Synthesized Data},
  author={Cheng Kang Chou and Chan-Jan Hsu and Ho-Lam Chung and Liang-Hsuan Tseng and Hsi-Chun Cheng and Yu-Kuan Fu and Kuan Po Huang and Hung-Yi Lee},
  journal={arXiv preprint arXiv:2506.11130},
  year={2025}
}
Downloads last month
25
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for phate334/Breeze-ASR-25-int8-CT2

Quantized
(15)
this model

Paper for phate334/Breeze-ASR-25-int8-CT2