Document Comparison
Most Diff Tools Show You Text.
DiffKit Shows You the Page.
Compare TXT, PDF, and DOCX files against each other — in Chinese, Arabic, Russian, or English — and see changes overlaid on the actual rendered document, not a wall of extracted text that lost the layout on the way out.
Replace the filter every six months.
Allow the adhesive strip to cure for at least one hour before applying load.
Every Format, Every Script, One Comparison
DiffKit doesn't ask you to convert anything first. Point it at two files — even two different formats — and it handles the rest.
Cross-Format Compare
TXT, PDF, and DOCX compared directly against each other — Master and Sample don't even need to be the same file type.
Real Multi-Script Support
Automatic Chinese, Japanese, Korean, Arabic (right-to-left), Thai, Devanagari, Cyrillic, and Latin tokenization — not just Unicode passthrough.
Visual PDF Compare
Dual, synchronized rendered pages with word-level change overlays, tap-to-jump navigation, and matched zoom between Master and Sample.
Formatting-Aware
Catches font, size, bold, and italic changes in PDF. DOCX gets the full set: family, size, weight, style, underline, strikethrough, and color.
Hard-PDF Repair
Recovers reading order across multi-column layouts and shattered glyph extraction, with an optional cloud fallback for scanned pages that need it.
Reports & Workflow
Export to PDF, HTML, or CSV, search inside a comparison, apply or undo edits, save zone profiles, and pick a session back up where you left it.
The Part Most Diff Tools Skip
Text-only diffing is easy. Showing you the same change, on the actual page, at the size it was printed — even when Master and Sample were reflowed into completely different layouts — is not.
Overlays sit on the real page
Word-level highlights render directly on the rendered PDF — not a synthetic reconstruction standing in for it.
Reading order, per column
A two-column reflow doesn't scramble Master and Sample into nonsense — each column is read top to bottom before moving to the next.
Zoom matches font size
Jumping to a change zooms Master and Sample to that change's true size — not a fixed zoom level — so a shrunk-down reprint still lines up visually with its original.
Or it tells you they can't align
When two layouts are too different to pair side by side at all, DiffKit switches to matching Sample to Master by meaning — and asks first, rather than guessing silently.
What DiffKit Actually Does
No hand-waving about what's "AI-powered." Here's the actual behavior.
- ✓Local-first extraction. Cloud processing is opt-in, limited to PDFs that need it, and never stores your files.
- ✓Exact source provenance. Every word-level change traces back to its precise position in the original document — not an approximation.
- ✓Layout mismatches, flagged. Dissimilar documents are detected automatically, and DiffKit asks before changing how it compares — never a silent switch.
- ✓Formatting, kept separate. A font change is reported as a font change, not mistaken for a content rewrite.
- ✓Script-correct diffing. Character-level comparison for CJK, word-level for everything else — because "word" doesn't mean the same thing in every script.
- ✓Built on real documents. Every comparison mode is validated against real Latin, Cyrillic, and CJK fixtures — not just English test strings.