I haven't really been able to get anyone to take note of this exploit yet, and don't personally have the time to do it all myself, so if anyone else wants to play around with this approach it 100% works on select portions of these documents. This is basic image manipulation for any graphic designers out there.
Dataset 7 contains documents from Epstein's first West Palm Beach investigation and grand jury testimonies.
These files contain scanned sheets of paper that were printed on both sides:
EFTA00009448.PDF
EFTA00009478.PDF
EFTA00009502.PDF
EFTA00009512.PDF
EFTA00009550.PDF
EFTA00009632.PDF
Within these documents, every page with a green hue can exploit the page that follows it. The green hue IS the text of the page that follows it.
If you take a green hue page, mirror it so that it reads backwards, then tease out the faint details (using the Levels tool in Photoshop, etc.), you can reveal numerous redacted words/names from the page that follows it.
If you then square up the mirrored green page on top of the next page (sometimes needing to adjust the angle, scale, etc.), you can then overlay a matching font which can either remove any doubt as to the redacted characters, or also assist when exploring less confidence fuzzy or obstructed portions of the text.
Every unredacted name I've found is public knowledge at this point, however it does provide indisputable proof surrounding some of the original testimonies and "co-conspirators" included the original plea deal immunity.