Orv PDF Translator
Upload a PDF or DOCX up to 40 MB, see the price before paying anything, and get back a print-ready translated document. Only the text goes to the translator; every other block — tables, figures, formulas — is reconstructed by the pipeline at its original coordinates, which is why the result still reads like the original instead of a re-flowed mess. Six subject domains, both directions detected automatically, first document free, files deleted after 72 hours.
// Architecture
A FastAPI service with a static UI served under pdftranslate.orvteam.com. Each page is decomposed with PyMuPDF and pdfplumber, the text blocks are translated, and the page is re-laid with ReportLab. Persian output is shaped with arabic-reshaper and python-bidi, wrapped intelligently, and auto-sized to fit its original box using self-hosted Persian fonts. Long jobs stream progress back, with a 40 MB upload ceiling and a 300-second timeout at the proxy.
// Key capabilities
Layout survives
Tables, images, charts and formulas are rebuilt at their original positions instead of being re-flowed into a wall of text.
Persian typeset properly
Right-to-left shaping, smart wrapping and adaptive font sizing over self-hosted Vazirmatn, Sahel and Samim.
Domain-specific modes
Scientific, medical, technical and legal modes change how terminology is handled instead of translating everything the same way.
Price before payment
The estimate is shown before any payment is taken, the first document is free, and uploaded files are deleted after 72 hours.
// Technology stack
- Python / FastAPI
- PyMuPDF
- pdfplumber
- ReportLab
- python-bidi
// Inside the project
// Next step
Need something like Orv PDF Translator?
Tell us what you're building. We'll reply with how we'd approach it, a rough timeline and the first step.
Start a similar project