Reglyph is our product, so we can't claim to be neutral. What we can do is make the test fair: same files, same target languages, default settings, and every result shown as it came out, flaws included.
| Page | Languages | Better result | What decided it |
|---|---|---|---|
| Text over a photo (scan) | EN → ZH | Reglyph | Background: Reflo replaced the photo with a solid backing |
| 1871 book page (scan) | FR → EN | Reglyph | Background: paper texture kept |
| Financial table (scan) | EN → JA | Reglyph | Coverage: Reflo left the table in English |
| Colored panels (scan) | EN → ES | Reglyph, slightly | Close: Reflo put a pale backing on the title bars and missed the footer; our tab label overflowed |
| Dense two-column paper (digital) | EN → FR | Reglyph | Typesetting: Reflo split words mid-word; we lost the indents |
| Paper with a parameter table (digital) | EN → KO | Tie | Layout even; Reflo translated code keys |
| Score | Reglyph 6 · Reflo 1 | A tie scores a point for each |
On three of the four scans, Reflo laid a solid-color backing under its translation. On five of the six pages, it left something untranslated.
Reglyph won all four scans, and scans are what we built it for. The first thing you notice is the background. Once the source text is erased, does the photo or paper come back, or is it replaced by a solid backing? That's the biggest gap between the two. Coverage is the other one: Reflo left whole sections untranslated on several pages.
The original is an aerial photo of a coastline with the English printed straight onto it. Reflo laid a solid gray-blue backing under every block of Chinese, so most of the sea on the left half of the page is no longer visible. On ours, the Chinese sits on the photo itself, with the image filled back in where the English used to be. We did get the three big numbers wrong (bold white in the original, thin and pale in ours), but this is the page where Reglyph's lead is widest.
This page comes from a real book printed in 1871, and you can see the age in the paper. Reflo set its English on a flat beige backing. It reads clearly, but the grain of the old paper doesn't come through. Neither tool kept the first-line indents, but the backgrounds aren't close: ours looks as if it had been printed on that same yellowed sheet.
On the financial report, Reglyph wins on coverage. Reflo translated the heading and the two paragraphs and stopped there, leaving the table, the chart labels, the header and the footer in English. On a financial statement, the table is usually the part you opened the file for. We translated close to everything and kept every number in its cell.
Of the four scans, this is the closest. Reflo's titles sit on a pale backing that interrupts the green gradient, the "Gut Health" tab comes out truncated as "Salud Intesti", and the footer stayed in English. We have one problem, with that same tab: the Spanish runs long, wraps onto two lines and pokes out of the box. Both sides have rough spots here, but Reflo's touch more of the page.
Digital PDFs already have clean text, usually on a white page, so putting backgrounds back barely comes into it. The hard part is getting longer text into the same space without breaking it. Reglyph handles a digital PDF the same way it handles a scan, and on these two papers the result held up.
The page was packed to begin with, and French runs longer than English. Both tools fit it in, but Reflo did it at a cost: several words are split across two lines in the middle ("o / nt", "i / ndiquent"), the journal name in the header runs together as "PHYSIQUEREVUELETTRES", and a few lines open with a stray underscore. Ours keeps every word whole and the equations in line with the text. Our one weakness here is that the paragraphs lost their first-line indents, which Reflo kept.
We scored this page a tie, since on layout there's nothing to choose between them. The difference is in the parameter table on the right. Reflo translated the configuration keys too, so n_layers became n_레이어, and if you search the model's code for that, you won't find it. We left the keys exactly as written.
More side-by-side tests: Reglyph vs DeepL on the same six pages, Reglyph vs Google Translate, Reglyph vs Smallpdf, and the best PDF translation services for a wider roundup. Not sure whether your file is a scan? Here's how to tell.
Upload a scanned PDF — even one made with your phone's scanner app. Reglyph OCRs it, translates it, and rebuilds the page so tables, stamps, and figures stay exactly where they were.
Translate 5 pages free, every dayIn our side-by-side test of four scanned pages, Reglyph kept the original background behind the translated text on all four: photos, colored title bars and old paper stayed visible. Reflo laid a solid backing under its translation on three of them, and left some text untranslated (tables, chart labels, captions, headers or footers) on five of the six test pages.
Yes. It read all four image-only pages in our test and translated the body text. On several of them it skipped table cells, chart labels, captions, headers and footers. On photos, colored bars and old paper, its approach is to lay a solid-color backing under the translation. That's fine for simple backgrounds; if you want the original background kept, it's not the right fit.
Reglyph, on all four scans we tried. It put the translation back onto photos, colored panels and paper texture instead of replacing them with a solid backing. It also translated the tables and captions Reflo skipped.
Reglyph did slightly better in our test. On a dense two-column paper going into French, Reflo split several words across lines and dropped the spaces in the header, while Reglyph kept the text whole but lost the paragraph indents. The second paper was a tie on layout; Reflo did translate code parameter names there, which Reglyph left alone.
We can't be neutral, so we tried to make it checkable instead. Both tools got the same files with default settings, every result is shown next to the original as it came out, and our own flaws are left in. All six source pages are public, so you can run the same test yourself.
Vision models genuinely read scans now. The question isn't whether ChatGPT can translate your PDF — it's whether a chat reply is actually what you need back.
Read guideCamera-translate apps are great for reading a page on the spot, but they don't give you a file — here's how to get an actual translated PDF from your phone.
Read guideRandom symbols, blank pages, or 'PDF tools are temporarily offline' — the cause is almost always the same one thing, and it's fixable.
Read guideSimple, scan-friendly pricing. Pages are pages — no multiplier for scanned files.
Translate a few pages. No commitment.
$0.30 a page, from 10 pages — you pay for exactly 10 pages.
For individuals with scans to translate every month.
For high-volume users with scans to translate every week.