1.3 KiB
1.3 KiB
| title | date | description | tags | |||
|---|---|---|---|---|---|---|
| Building a PDF Translation Pipeline | 2026-02-11 | How I built a pipeline to translate Hindi PDFs to English using OCR and neural machine translation. |
|
Recently, I needed to translate a PDF document from Hindi to English. Sounds simple enough, right? Turns out, it's a surprisingly deep rabbit hole.
The Pipeline
The approach I settled on follows this flow:
PDF → Images → OCR → Translation → Rendered Images → PDF
Each step has its own set of challenges:
- PDF to Images: Convert each page to a high-DPI image for better OCR accuracy
- OCR: Extract text with position data using PaddleOCR
- Translation: Run extracted text through NLLB (No Language Left Behind)
- Rendering: Paint translated text back onto the original image
- Assembly: Combine rendered images back into a PDF
Lessons Learned
- DPI matters a lot — bumping from 150 to 300 DPI dramatically improved OCR accuracy for Hindi text
- Font rendering is hard — getting translated text to fit in the same bounding boxes required careful font size calculation
- Fallback strategies — TrOCR as a fallback when PaddleOCR fails on certain text regions
The code is messy but it works. Sometimes that's enough.