Video Transcription Tool

A command-line utility that converts video files to readable transcripts and summaries using OpenAI's APIs.

Features

• Extracts optimized audio from video files using FFmpeg
• Transcribes audio to text using OpenAI's Whisper API
• Formats transcripts into readable HTML with proper structure
• Generates concise summaries of video content
• Handles batch processing of multiple videos
• Intelligently caches intermediate files to avoid redundant processing

Requirements

• Python 3.6+
• FFmpeg installed and available in PATH
• OpenAI API key

Installation

1. Clone this repository:

   git clone https://github.com/iliakan/transcript.git
   cd transcript

2. Install required Python packages:

   pip install requests

3. Set your OpenAI API key as an environment variable:

   export OPENAI_API_KEY="your-api-key-here"

Usage

Process a single video

python video_to_text.py /path/to/your/video.mp4

Process all videos in the 'video' directory

python video_to_text.py

How It Works

1. Audio Extraction: The tool extracts audio from the video file, optimizing it for transcription with appropriate bitrate and filtering.

2. Transcription: The audio is sent to OpenAI's Whisper API for transcription.

3. Formatting: The raw transcript is processed by GPT-4o to create a well-formatted HTML version with proper paragraphs and sections.

4. Summarization: GPT-4o generates a concise summary of the video content in HTML format.

5. Output: Both the formatted transcript and summary are saved in the project directory structure.

Output Directory Structure

The tool organizes processed files in a flat structure with separate directories for each file type:

├── audio/
│   └── video_name.mp3       # Extracted audio files
├── transcript/
│   ├── video_name.txt       # Raw VTT transcripts
│   └── video_name.html      # Formatted HTML transcripts
├── summary/
│   └── video_name.html      # HTML summaries
└── video/
    └── video_name.mp4       # Source video files

Notes

• The tool optimizes audio extraction to keep files under 25MB (OpenAI's file size limit)
• All files (audio, transcripts, HTML, and summaries) are cached to avoid redundant processing
• Place your video files in the 'video' directory for batch processing
• Detailed error information is displayed for OpenAI API errors to help with troubleshooting

License

MIT License