Video Supercut

This project processes videos by extracting audio, detecting speech segments, transcribing with Whisper, removing duplicate phrases, and compiling a final edited video.

Prerequisites

• Python 3.6 or higher
• ffmpeg must be installed and available in your system PATH (the script will try to use imageio-ffmpeg if you haven't read this).
• Git for cloning the repository.

Setup

1. Clone the repository:

   git clone git@github.com:justinkahrs/video-supercut.git

2. Navigate into the project directory:

   cd video-supercut

3. Create and activate a virtual environment:

   python3 -m venv venv
   source ./venv/bin/activate

4. Install dependencies:

   pip install -r requirements.txt

Configuration Parameters

• NOISE: The noise level environment variable (low, medium, or high). The script measures the audio's average RMS in dBFS and adjusts the silence threshold accordingly.
• MIN_SPEECH_DURATION: The minimum duration (in milliseconds) that a segment must last to be considered speech.
• MARGIN_DURATION: An extra margin (in milliseconds) added before and after detected speech segments to ensure that speech is fully captured.
• RAW_FOLDER: The directory where raw .mp4 video files should be placed.
• EDITED_FOLDER: The directory where the final edited videos will be saved.

Running the Script

1. Add Video Files:

   • mkdir raw
   • Place your .mp4 video files in the raw folder.

2. Run the Script:

   python3 main.py

   The script will process the videos and output the edited videos to the edited_videos folder.

Notes

• The first run may download the necessary Whisper model weights.
• Temporary files are stored in the system's temporary directory.
• Ensure that ffmpeg is installed and configured correctly on your system.