Disclaimer: Unredact is a research and entertainment project — not a forensic tool. Nothing it produces should be taken as verified fact. All data comes from open-source dictionaries and publicly available font metrics. AI scores reflect plausibility, not truth. Do not use this tool to circumvent lawful redactions or violate anyone's privacy. Use responsibly and in accordance with applicable laws. The creator accepts no liability for misuse.

This video is about a problem that has been considered unsolvable: the permanent redaction of the Epstein case files. As an engineer, I don't see a black bar as a dead end—I see it as a geometry problem with a data-driven solution.

In this video, I demonstrate Unredact, a tool I built to reverse-engineer these redactions using a multi-phase algorithm that layers OCR, Computer Vision, and AI. By cross-referencing the physical dimensions of the redaction boxes against known typography and massive datasets, we can move past speculation and into forensic reality.

The goal of this project is simple: accountability. If the data exists to uncover the truth, that data should be accessible to the public. To that end, Unredact is now fully open-source.

Credits & Resources
This project was built on the shoulders of these incredible investigators:

Inspiration: Huge thanks to @EpsteinSleuther. Her methodology and deep-dives into these files provided the blueprint for how I structured this application’s detection logic.

Channel: https://www.youtube.com/@EpsteinSleuther

Data Source: This tool wouldn't be possible without the work of R. Howard Stone, whose research and associate data provided the essential validation sets for the algorithm.

Data Repo: https://github.com/rhowardstone/Epstein-research-data

Get the Tool
You can find the code, the logic breakdown, and the contribution guidelines on GitHub. I’m making this available to everyone so we can collectively ensure these names don't stay hidden forever.

Project Repo: https://github.com/Alex-Gilbert/unredact

#EpsteinFiles #Forensics #OpenSource #DataScience #Unredact

YouTube

A developer presents Unredact, an open-source tool that combines OCR, computer vision, and AI to attempt to identify names and words hidden behind redaction bars in the publicly released Epstein court documents. The tool analyzes surrounding typography, measures pixel dimensions of redaction boxes, detects pixel bleed artifacts at box edges, and cross-references a database of known Epstein associates to generate ranked candidate names. An LLM validation feature narrows down candidates by using surrounding sentence context. The project is released as open source for public contribution.

Unmasking Epstein: My Open Source AI Tool