Quill

A command-line tool for transcribing audio files, YouTube videos, and podcasts using Faster Whisper.

Prerequisites

• ffmpeg
• Python 3.10+

Installation

At system level you need cuda 12 and cuddn.

sudo apt install libcublas12
sudo apt install cudnn9-cuda-12

# Install dependencies and create virtual environment
uv sync

# Optional: Install globally as a CLI tool
uv tool install . --python 3.12

Usage

uv run quill INPUT_SOURCE OUTPUT_FILE [--model MODEL] [--device DEVICE] [--language LANGUAGE]

# Examples:
uv run quill audio.mp3 transcript.txt
uv run quill https://youtube.com/watch?v=... transcript.txt
uv run quill https://example.com/audio.mp3 transcript.txt --model large
uv run quill podcast.mp3 output.txt --device cuda --language en

Options:

• --model: Model size to use (tiny, base, small, medium, large). Default: medium
• --device: Device to use for inference (cpu, cuda). Default: cpu
• --language: Language code for transcription (e.g., en, fr, de). Default: auto-detect
• --timestamps, -t: Include timestamps in the transcription output

Supported Input Sources

• Local audio files (mp3, wav, m4a, etc.)
• YouTube URLs
• Direct URLs to audio files

Performance Notes

• Using CUDA-enabled GPU significantly improves transcription speed
• Larger models provide better accuracy but require more memory and processing time
• The 'medium' model provides a good balance between speed and accuracy for most use cases

License

Apache 2.0