Views
No views yet
pip install faster-whisper1from faster_whisper import WhisperModel
2
3model = WhisperModel(
4 "LumyAgency/bretagne-whisper-ct2",
5 device="cpu",
6 compute_type="int8"
7)
8
9segments, info = model.transcribe("audio.mp3", language="br")
10
11for segment in segments:
12 print(f"[{segment.start:.2f}s] {segment.text}")1from faster_whisper import WhisperModel
2
3model = WhisperModel(
4 "LumyAgency/bretagne-whisper-ct2",
5 device="cuda",
6 compute_type="float16"
7)
8
9segments, info = model.transcribe("audio.mp3", language="br")
10
11for segment in segments:
12 print(f"[{segment.start:.2f}s] {segment.text}")1import whisperx
2
3model = whisperx.load_model(
4 "LumyAgency/bretagne-whisper-ct2",
5 device="cuda",
6 compute_type="float16",
7 language="br"
8)
9
10audio = whisperx.load_audio("audio.mp3")
11result = model.transcribe(audio)
12print(result["text"])1curl -X POST "http://votre-serveur:8000/service/transcribe" \
2 -F "file=@audio.mp3" \
3 -F "language=br" \
4 -F "model=LumyAgency/bretagne-whisper-ct2" \
5 -F "device=cuda" \
6 -F "compute_type=float16"| Paramètre | Valeur |
|---|---|
| Architecture | Whisper Large-v3-Turbo |
| Paramètres | 809M |
| Encoder layers | 32 |
| Decoder layers | 4 |
| Quantization | float16 |
| Taille | ~1.5 GB |
| Format | CTranslate2 |
1@misc{bretagne-whisper-ct2,
2 title={Bretagne Whisper Large-v3-Turbo CTranslate2},
3 author={Conversion par LumyAgency},
4 year={2025},
5 url={https://huggingface.co/LumyAgency/bretagne-whisper-ct2}
6}