CHAOS
Oct 2026
Counter-Strike Highlight Analysis and Organization System. Point it at hours of gameplay footage and it finds the moments worth keeping (kills, funny voice chat, good conversations) and clips them for you. Runs entirely on your own machine.
Why I Built It
I have 700+ GB of my own Counter-Strike footage from 10+ years of playing and making videos, and scrubbing through it by hand was never going to happen. The rewrite made it a lot faster and more accurate. Classical computer vision does most of the work and OCR only runs on small crops. It samples every couple of seconds, then binary searches for the exact frame of a kill, which was the big speed win. An audio tagger catches laughter in voice chat, Whisper adds burned-in subtitles, and a resource planner sizes the workers from each model's measured memory so it doesn't eat my whole computer. It can resume where it left off, recompute only what changed, and throttle itself while I'm gaming. No LLMs, no cloud calls.
Technologies
- Python
- PyQt6
- OpenCV
- PyTorch
- Whisper
- FFmpeg
Tags
- Machine Learning
- Computer Vision
- OCR
- Speech-to-Text
- Audio Processing
- Video Processing
- Batch Processing
- Open Source
Links
- Visit site: https://github.com/nick5616/CHAOS
- Source: https://github.com/nick5616/CHAOS