I highlight paper books. A small open model types my quotes while I sleep

Description and use

Dogear automatically extracts highlighted passages from paper books using OCR and the Gemma 4 model, presenting them for review the next morning and saving hours of manual transcription.

Problem

Copying highlighted passages from paper books into a digital format is time‑consuming and tedious.

Outcome

Dogear reduces manual transcription from 30–60 minutes per book to a single photo per highlighted page and a brief morning review, costing only a few cents per night.

Workflow

• Add a book by typing or photographing its cover
• Photograph each highlighted page while the book is open
• Upload photos to a private bucket
• Nightly cron job runs Tesseract OCR and Gemma 4 on a CPU to extract highlighted passages
• Store each quote with a crop of the passage
• Delete processed photos
• Morning review: keep, edit, delete or add missed quotes
• Export to Notion (planned)

Evidence

Dogear removes that chore. While reading, the only extra thing I do is take a photo of each page I highlighted. Overnight, a small open-weight model running on a plain CPU in the cloud reads the photos and pulls out exactly the highlighted passages.

Source

The night-run numbers are a real line from the Render log.

Source

The full loop works: library, page capture, night run and morning review.

Source