AI Transcript

AI-powered video transcription with Whisper and VAD for accurate subtitle generation

Generate high-quality subtitles from video or audio files using OpenAI's Whisper model with intelligent voice activity detection (VAD) for optimal accuracy.

Features

• 🎯 High Accuracy: Combines Silero VAD + Faster-Whisper for precise speech detection and transcription
• ⚡ GPU Accelerated: CUDA support with 2.8x faster processing (GTX 1080 tested)
• 📝 Multiple Formats: Export to SRT or VTT subtitle formats
• 🎬 Video & Audio: Supports MP4, AVI, MKV, MOV, WAV, MP3, FLAC, and more
• 🛠️ Flexible Models: Choose from tiny to large Whisper models based on accuracy/speed needs
• 🌍 Multi-language: Supports 99+ languages including English, Chinese, Spanish, French, etc.
• 💻 Clean CLI: Professional command-line interface with progress indicators
• 🧪 Well Tested: 206 tests with comprehensive coverage

Installation

Prerequisites

• Python 3.9 or higher
• uv (recommended) or pip
• FFmpeg (for video processing)
• CUDA-capable GPU (optional, for acceleration)

Install with uv (Recommended)

# Clone repository
git clone https://github.com/TMYuan/ai-transcript.git
cd ai-transcript

# Install dependencies
uv sync

# Install FFmpeg (if not already installed)
# Ubuntu/Debian:
sudo apt install ffmpeg
# macOS:
brew install ffmpeg

Install with pip

git clone https://github.com/TMYuan/ai-transcript.git
cd ai-transcript
pip install -e .

Quick Start

# Basic usage (auto-detects GPU)
uv run aitranscript transcribe video.mp4

# Specify model size and device
uv run aitranscript transcribe video.mp4 --model medium --device cuda

# Generate VTT format with custom output
uv run aitranscript transcribe audio.wav --format vtt -o subtitles.vtt

# Transcribe in Chinese
uv run aitranscript transcribe video.mp4 --language zh

CLI Usage

Transcribe Command

uv run aitranscript transcribe [OPTIONS] INPUT_FILE

Options:

Option	Description	Default
-o, --output PATH	Output subtitle file path	Auto-generated
--model MODEL	Whisper model size: tiny, base, small, medium, large	medium
--device DEVICE	Compute device: cuda, cpu	cuda
--format FORMAT	Subtitle format: srt, vtt	srt
--language CODE	Language code (en, zh, es, fr, etc.)	en
-q, --quiet	Minimal output (no progress bars)	-
-v, --verbose	Detailed logging for debugging	-

Examples:

# Fast transcription with tiny model on CPU
uv run aitranscript transcribe video.mp4 --model tiny --device cpu

# High accuracy with large model on GPU
uv run aitranscript transcribe video.mp4 --model large --device cuda

# Quiet mode for scripting
uv run aitranscript transcribe video.mp4 --quiet -o output.srt

# Verbose output for debugging
uv run aitranscript transcribe video.mp4 --verbose

Help

# Show help
uv run aitranscript --help

# Show transcribe command help
uv run aitranscript transcribe --help

# Show version
uv run aitranscript --version

Supported Formats

Video: MP4, AVI, MKV, MOV, FLV, WMV, WebM Audio: WAV, MP3, FLAC, AAC, OGG, M4A

Project Structure

ai-transcript/
├── src/aitranscript/
│   ├── cli/
│   │   └── main.py              # Click-based CLI interface
│   ├── core/
│   │   ├── audio_extractor.py   # Video → Audio extraction
│   │   ├── transcriber.py       # Whisper transcription
│   │   ├── vad_processor.py     # Speech detection with Silero VAD
│   │   └── subtitle_generator.py # SRT/VTT generation
│   ├── models/
│   │   ├── config.py            # Configuration dataclasses
│   │   └── segment.py           # Speech/Transcript segments
│   ├── pipeline/
│   │   └── transcription_pipeline.py # End-to-end orchestration
│   └── utils/
│       ├── file_utils.py        # File operations
│       ├── validators.py        # Input/output validation
│       └── logger.py            # Centralized logging
├── tests/                       # Comprehensive test suite (206 tests)
├── pyproject.toml               # Project configuration
└── README.md                    # This file

License

MIT License - see LICENSE file for details.

Acknowledgments

• OpenAI Whisper - Speech recognition model
• Faster-Whisper - Optimized Whisper implementation
• Silero VAD - Voice activity detection
• Click - CLI framework
• Rich - Terminal formatting

Support

• Issues: GitHub Issues
• Discussions: GitHub Discussions