VideoHighlighter (Freeware)
A Python tool to automatically generate highlight clips from videos using scene detection, motion detection, audio peaks, object detection, action recognition, and transcript analysis.
Features
Detects:
- Scenes using OpenCV.
- Motion peaks and scene changes.
- Objects
- Actions
- Audio peaks.
Generates transcript subtitles via OpenAI Whisper. Cuts and merges top scoring segments into a highlight video. Fully configurable: frame skip, highlight duration, keywords. Optional GUI for easy interaction.
Setup & Installation
- Python & FFmpeg FFmpeg must be installed and available in your system PATH.
Check FFmpeg installation: ffmpeg -version
- Install Python Dependencies pip install -r requirements.txt
Dependencies include: numpy, torch, opencv-python, tqdm, ffmpeg-python, openai-whisper, googletrans, openvino-dev For YOLO object detection: ultralytics For GUI: PySide6
- Download Models The code will automatically download YOLO models on first run.
Usage Linux: python main.py Windows: run Videohighlighter.exe
Notes
OpenAI Whisper is MIT licensed — freely usable.
Google Translate API is optional. If using unofficial libraries (googletrans), no API key is needed, but results may break if Google changes endpoints.
This project does not include any paid API keys. Users must provide their own if using official services.
License
This repository is released under the GNU Affero General Public License v3.0 (AGPLv3). You are free to use, modify, and distribute the code, provided that any modified versions, including those offered over a network, make their complete source code available under the same license.
Project Background
This project started as a personal tool to automatically generate subtitles for videos, for my young 7 years old son. Over time, it evolved into a highlights generator for movies, sports, and personal videos.
The primary goal remains practical: speed up video analysis, generate highlights, and create accessible subtitles automatically.
