Reglyph translates scanned and image-only PDFs while keeping the original layout. OCR reads the page, the source text is erased, and the translation is re-typeset in place — tables, columns, stamps and numbers stay put. 5 pages free.
No reviews yetBe the first to leave a review for Reglyph
Maker
📌
Hi Product Hunt! 👋
I built Reglyph because every "PDF translator" I tried fell apart on scanned documents — Google Translate says "no text found", ChatGPT silently reads nothing, and the tools that do work dump your beautifully formatted contract into a wall of plain text.
Reglyph treats scanned, image-only PDFs as the primary use case:
🔍 OCR reads the page (no pre-conversion needed)
🧽 The original text is erased from the image with inpainting
✍️ The translation is re-typeset in the exact same position
So tables, columns, figures, stamps, signatures and numbers stay exactly where they were, from certificates, contracts, textbooks to whole scanned books.
You can also export a side-by-side bilingual PDF for proofreading.
Try it free (5 pages, no credit card): https://tryreglyph.com
I'd love to hear what documents break it, every weird scan makes it better!
What's the worst PDF you've ever had to translate? 😄
Report
Keeping the page intact while translating documents sounds deceptively difficult. Curious whether preserving formatting or preserving meaning has turned out to be the harder engineering challenge in practice.
Report
Maker
@aryan787544 Formatting, by a wide margin 🙂 Translation quality mostly comes free with modern models — the hard part is layout: deciding which text lines belong together (table rows love to merge), erasing the original cleanly, and re-typesetting text that got 40% longer into the same box. Great question!
Report
@li_lu5That's an interesting inversion. A few years ago everyone assumed translation was the hard problem. Now it sounds like understanding the document's structure has become the real challenge.
I'll be curious whether users eventually choose Reglyph because it's the most accurate translator, or because it's the one they trust with documents they can't afford to manually fix afterward. Those feel like very different long-term positions.
Report
finally something that handles scanned PDFs without mangling the layout. threw a Japanese contract at it and the table headers actually lined up after translation
Report
Maker
@narinengnejqio That's exactly the test case we built this for — thanks for trying it! If you ever hit a table that doesn't line up, send it over, those make the best test files.
Report
Finally something that doesn't butcher tables and stamps when translating scanned PDFs. Tried a German contract and the columns actually lined up properly, which feels almost magical after years of fighting clunky OCR tools.
Report
Maker
@dorukkzltajeny "Almost magical" is going on the fridge! 😄 Thanks for testing it on a real contract — German's long compound words are a fun stress test for the re-typesetting.
Report
Finally a tool that keeps my Japanese technical manuals readable after translation, the columns and stamps stayed aligned which I was not expecting. Worth the free five pages just to see it work.
Report
Maker
@rseybek97138 Technical manuals are one of the toughest layouts out there, so this is great to hear. Thanks for giving the free pages a spin!
Keeping the page intact while translating documents sounds deceptively difficult. Curious whether preserving formatting or preserving meaning has turned out to be the harder engineering challenge in practice.
@aryan787544 Formatting, by a wide margin 🙂 Translation quality mostly comes free with modern models — the hard part is layout: deciding which text lines belong together (table rows love to merge), erasing the original cleanly, and re-typesetting text that got 40% longer into the same box. Great question!
@li_lu5That's an interesting inversion. A few years ago everyone assumed translation was the hard problem. Now it sounds like understanding the document's structure has become the real challenge.
I'll be curious whether users eventually choose Reglyph because it's the most accurate translator, or because it's the one they trust with documents they can't afford to manually fix afterward. Those feel like very different long-term positions.
finally something that handles scanned PDFs without mangling the layout. threw a Japanese contract at it and the table headers actually lined up after translation
@narinengnejqio That's exactly the test case we built this for — thanks for trying it! If you ever hit a table that doesn't line up, send it over, those make the best test files.
Finally something that doesn't butcher tables and stamps when translating scanned PDFs. Tried a German contract and the columns actually lined up properly, which feels almost magical after years of fighting clunky OCR tools.
@dorukkzltajeny "Almost magical" is going on the fridge! 😄 Thanks for testing it on a real contract — German's long compound words are a fun stress test for the re-typesetting.
Finally a tool that keeps my Japanese technical manuals readable after translation, the columns and stamps stayed aligned which I was not expecting. Worth the free five pages just to see it work.
@rseybek97138 Technical manuals are one of the toughest layouts out there, so this is great to hear. Thanks for giving the free pages a spin!