Software · 2026

Local OCR pipeline

Turns scanned PDFs into searchable text on my own machine, using a vision model in LM Studio.

Starter text. Replace with the real story.

I wanted searchable text from scanned PDFs without sending the files to a cloud service. The pipeline runs a vision-language model through LM Studio on an RTX 5090 and writes the text back out.

How it works

  1. Split each PDF into page images
  2. Send each page to the local model
  3. Collect the text and save it alongside the original

Results

Add the numbers here: pages per minute, accuracy on a test set, anything you measured.

All projects