whisper-turbo-tedlium-ct2

Whisper Turbo Fine-tuned on TED-LIUM (CTranslate2 Format)

This is the CTranslate2 version of Dafisns/whisper-turbo-tedlium, optimized for fast inference with faster-whisper.

🚀 Performance

• 2-4x faster than original PyTorch model
• Lower memory usage
• Optimized for GPU inference

📦 Installation

pip install faster-whisper

💻 Usage

from faster_whisper import WhisperModel

# Load model (auto-download from HuggingFace)
model = WhisperModel("Dafisns/whisper-turbo-tedlium-ct2", device="cuda", compute_type="float32")

# Transcribe
segments, info = model.transcribe("audio.mp3", language="en")

for segment in segments:
    print(f"[{segment.start:.2f}s -> {segment.end:.2f}s] {segment.text}")

Model Info

• Base Model: OpenAI Whisper Large v3 Turbo
• Fine-tuned on: TED-LIUM Release 3
• Format: CTranslate2 (optimized)
• Quantization: float32
• Language: English

Related Models

• PyTorch Version: Dafisns/whisper-turbo-tedlium
  • Use this for training/fine-tuning
• CTranslate2 Version: Dafisns/whisper-turbo-tedlium-ct2
  • Use this for fast inference

Citation

@misc{whisper-turbo-tedlium-ct2,
  author = {Dafisns},
  title = {Whisper Turbo Fine-tuned on TED-LIUM (CTranslate2)},
  year = {2024},
  publisher = {HuggingFace},
  url = {https://huggingface.co/Dafisns/whisper-turbo-tedlium-ct2}
}

📄 License

Apache 2.0