Read an upside-down, blurry page scan with Indra OCR on a CPU

Try it yourself

What you get

Clean markdown text from a scanned page, even one that is upside down, blurry and grainy, with no GPU.

Tools

• Indra OCR (Datalab)
• Any recent CPU; we used an 8-vCPU Intel Xeon

Time / cost: 0.46 s of OCR, 1.1 s for the whole command (model already downloaded); 6.2 s for a 56-page paper, every page OCR'd

Steps

1. Install it with pip install indra-ocr (about 10 seconds in our test).
2. Run indra-ocr on your scan or PDF with --format md; add --timings to see how long each page takes.
3. For a PDF that already has a text layer, the default mode uses it; --mode ocr forces every page through OCR.
4. Check superscripts, footnote marks and sideways text by hand: those came out garbled in our test.

Watch out: Code Apache-2.0. Weights: modified OpenRAIL-M, free for personal use, research, and organizations at or under $20M in revenue or funding; not for anyone offering a product that competes with Datalab's. No region limits.

Proof / source

Tested by AgentPlays Our own test on Oct 15, 2026, on an 8-vCPU Intel Xeon, CPU only: input scan, output text and timings. Model card on Hugging Face, code on GitHub.