capitaine/fast-whisper-api repository overview

⁠FastWhisper API Base Image

Production-ready base image for speech-to-text transcription API using faster-whisper (SYSTRAN) with CUDA acceleration.

⁠What's Included

• PyTorch 2.6 with CUDA 12.4 and cuDNN 9
• faster-whisper - Fast Whisper inference engine
• FastAPI & Uvicorn - API framework and server
• ffmpeg - Audio processing (MP3, WAV, etc.)
• Build tools - ninja, build-essential for extensions

⁠Quick Start

FROM capitaine/fast-whisper-api:latest

# Copy your application code
COPY app/ ./app/

# Set environment variables
ENV DEVICE_ID=0
ENV BATCH_SIZE=24

EXPOSE 8000

CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "8000"]

⁠Features

• ✅ GPU-accelerated transcription with NVIDIA CUDA
• ✅ Multi-language support (98+ languages)
• ✅ Translation capabilities
• ✅ Optimized for production deployments
• ✅ Pre-installed dependencies for fast builds

⁠Requirements

• Docker with NVIDIA GPU support
• NVIDIA Container Toolkit
• CUDA-compatible GPU

⁠Source Code

GitHub: https://github.com/capisoft-lib/FastWhisperApi⁠

⁠License

MIT License