Read the PDF in your language, on the same page.
Press Translate in the PDF toolbar and pick a language. The text changes where it sits. Equations, code, numbers and links stay as they were.
Get started3. Results
The model reached 94.1% accuracy on the test set, 2.3 points above the baseline.
L = −Σ yᵢ log ŷᵢ
Code and data: www.lab.org/study
3. Ergebnisse
Das Modell erreichte 94.1% Genauigkeit auf dem Testdatensatz, 2.3 Punkte über der Baseline.
L = −Σ yᵢ log ŷᵢ
Code und Daten: www.lab.org/study
From the toolbar to a new PDF.
The words change. The page does not.
Text fits its box
Each translation is drawn inside the box the original text used. If it runs long, the type shrinks evenly, down to 60 percent. If it still does not fit, the app flags it as Overflow so you can fix it.
Maths and code are left alone
The app never sends blocks it reads as equations, formulas, code or figures. Inside the text, numbers, URLs, email addresses and inline code are locked. If one comes back changed, the app rejects that section.
The PDF's own fonts
The translation uses the fonts embedded in the PDF where they have the characters. Where they do not, Noto Sans, Noto Serif or Noto Sans JP stands in, and the review list says which.
Scans work too
A scanned PDF has no text to translate yet. Oyren offers to make the text editable first, then carries on with the translation. Both steps use AI credits.
27 languages, read left to right.
- Azerbaijani
- Chinese (Simplified)
- Chinese (Traditional)
- Czech
- Danish
- Dutch
- English
- Finnish
- French
- German
- Greek
- Hungarian
- Indonesian
- Italian
- Japanese
- Korean
- Norwegian
- Polish
- Portuguese
- Romanian
- Russian
- Slovak
- Spanish
- Swedish
- Turkish
- Ukrainian
- Vietnamese
Arabic and Hebrew are not available yet. Right-to-left layout is not supported, so the app leaves them switched off rather than getting the page wrong.
Translation runs on AI credits, the same credits as the rest of Oyren. There is no subscription.