42-evey/ghost-font-breaker — explained in plain English
Analysis updated 2026-05-18
Recover a hidden message from a Ghost Font video file or URL.
Extract and save the decoded message, decoy text, and motion report as image and text files.
Test the decoder against provided example video clips before trying your own.
| 42-evey/ghost-font-breaker | 0marildo/imago | 0xdfi/glm-5.2-1m-4x-dgx-spark | |
|---|---|---|---|
| Stars | 3 | 3 | 3 |
| Language | Python | Python | Python |
| Setup difficulty | easy | easy | hard |
| Complexity | 3/5 | 2/5 | 5/5 |
| Audience | developer | general | researcher |
Figures from each repo's GitHub metadata at analysis time.
Claude-based OCR is optional and needs an API key, the offline RapidOCR path works without one.
Ghost Font Breaker is a tool that decodes hidden messages from videos made with Ghost Font, a font designed to hide text from AI by writing it in the motion of dots instead of in any single picture frame. If you just look at one frame, or even average many frames together, you only see random noise or a fake decoy message. The real message only appears once you account for the motion of the dots. The tool works by tracking how one group of dots moves in a different direction than the background dots. Once it freezes the moving group in place, the hidden letters become visible while the background blurs into nothing. You point the tool at a Ghost Font video, either a local file or a URL, and it prints out the recovered motion details, the decoy text, and the actual hidden message. It also saves several image files to a folder, including the decoded message, an inverted version, the decoy image, and a full text report. After the motion is recovered, the tool still needs to read the text out of a somewhat fuzzy image. For this it can use Claude's vision model if you provide an API key, which the README describes as very accurate for this task. If no key is set, it falls back to RapidOCR, an offline text recognition tool that requires no extra system software, only Python packages. If neither option is available, the extracted image is still saved so a person can read it directly. Setup is done through simple scripts for macOS, Linux, or Windows, and the only requirements are a few Python packages such as NumPy, OpenCV, and RapidOCR, with no other system tools needed. The project includes example video clips you can download and test with a single command. According to the README, both provided example clips decode correctly, including one that had been recompressed after being posted to Twitter. The project is released under the MIT license and welcomes outside contributions.
A tool that decodes secret text hidden in Ghost Font videos by tracking the motion of scrolling dots.
Mainly Python. The stack also includes Python, NumPy, OpenCV.
Use freely for any purpose, including commercial use, as long as you keep the copyright notice.
Setup difficulty is rated easy, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.