vid2text is a powerful command-line interface (CLI) tool designed for extracting searchable transcriptions from a variety of video sources, including YouTube videos, local files (.mp4, .avi, .mov, .mkv, .m4v), and M3U8 streams. Utilizing innovative local Whisper models, this tool allows for efficient and effective transcription.

Key Features

• Multi-Source Support: Seamlessly transcribe videos from YouTube, local storage, and M3U8 streams.
• Local Transcription Capabilities: Leverage MLX Whisper for macOS or the cross-platform OpenAI Whisper to process videos.
• SQLite Database Integration: Store transcriptions in a searchable SQLite database with a user-friendly Datasette web interface.
• Batch Processing: Configure multiple videos through a simple YAML file, making it easy to handle extensive transcription tasks.

Usage Examples

Single Video Transcription

Process a video with the following commands:

vid2text youtube "https://youtu.be/VIDEO_ID"
vid2text local "/path/to/video.mp4"
vid2text m3u8 "https://example.com/stream.m3u8"

# Use options for more control
vid2text --model small.en --verbose youtube "https://youtu.be/..."
vid2text --dry-run local video.mp4  # Preview the operation

Batch Processing Setup

Create a config.yaml file with video settings:

videos:
  youtube:
    - url: "https://youtu.be/dQw4w9WgXcQ"
    - url: "https://youtu.be/jNQXAC9IVRw"
      title: "Custom Title"  # Optional
  local:
    - path: "/path/to/video.mp4"
    - path: "/path/to/folder/"  # Processes all videos in the folder
      title: "Folder Videos"
  m3u8:
    - url: "https://example.com/video.m3u8"
      title: "Live Stream"
      order: 1

settings:  # Optional
  whisper_model: "small.en"  # Overrides the default model
  log_level: "DEBUG"

Process all videos listed in the configuration:

vid2text process config.yaml
vid2text --dry-run process config.yaml  # Preview before processing

Database Operations

Access various database functionalities:

vid2text stats                           # Display video count
vid2text --db-path custom.db stats       # Use a custom database
vid2text view                            # Launch the web interface (requires Datasette)
vid2text view --port 8080                # Specify a custom port

Configuration Options

Configure the environment variables for enhanced control over the tool's operations:

Whisper Models Available

For macOS (Apple Silicon) - MLX Whisper Models:

• mlx-community/whisper-medium.en-mlx (default)
• mlx-community/whisper-large-v3-mlx
• mlx-community/whisper-small.en-mlx

Cross-Platform Models - OpenAI Whisper:

• base.en (default)
• tiny.en
• small.en
• medium.en
• large

Common Commands Reference

• vid2text youtube <url> - Transcribe a YouTube video.
• vid2text local <path> - Process a local video or folder.
• vid2text m3u8 <url> - Handle M3U8 stream transcriptions.
• vid2text process <config.yaml> - Execute batch processing based on your YAML configuration.
• vid2text stats - Check statistics on transcriptions stored in the database.
• vid2text view [--port PORT] - Open the Datasette web interface.

Support and Troubleshooting

For issues such as FFmpeg installation or memory-related problems during transcription, solutions and guidance are available in the README documentation.