Skip to content

A scanned PDF comes back garbled

Short answer

Scanned pages are read by a vision model rather than by OCR, so poor scans, skewed pages and handwriting degrade badly. Only the first 20 pages are processed.

Why this happens

There is no OCR engine in the pipeline. A model looks at the page image and reads it, which handles layout well and handles bad scans poorly.

What to do

  1. 01Use a vision-capable model: Qwen 3.5 or Gemma 4.
  2. 02If the scan is skewed or low-contrast, re-scan or straighten it. Image quality maps directly onto accuracy here.
  3. 03Split documents longer than 20 pages and ask about each part.
  4. 04For anything where exactness matters — figures, names, dates — check them against the page rather than trusting the transcription.

Other ways people ask this

  • nightjar scanned pdf comes back garbled
  • Nightjar Browser: scanned pdf comes back garbled. How do I fix it?

Related