Blog

Benchmarks, know-how for hard-to-read documents, and what's new in Runemic.

How well does OCR read Persian? We tested six engines

On eight Persian test pages, the best engine got 3.1% of characters wrong and the worst 18.5%. Here is how we tested, what each engine got wrong, and how to rerun it.

Read Know-how

Why OCR struggles with Persian, Arabic and Urdu

Connected letters, dots that change the letter, calligraphic styles and mixed directions. What makes Arabic-script text hard for OCR, and how to get better results from any engine.

Read Product update

Four open OCR models, now in the console

GLM-OCR, LightOnOCR-3, dots.mocr and KDL Frontier Parser nano now run on Runemic's own GPUs. What each is good at, what they cost, and why the first page can take a few minutes.

Read

Try Runemic OCR free

Read your own pages in the browser, with no account. When you want more, sign in with GitHub or Hugging Face.

Free at runemic.com/try, no sign-up
Sign in for 2,500 free credits, no card needed
Compare up to three models, and get an API key with cURL, Python and JavaScript examples

Try it free Get an API key

# one page in, Markdown out
curl https://api.runemic.com/v1/ocr \
  -H "Authorization: Bearer $RUNEMIC_KEY" \
  -F "file=@page.png"

# → { "text": "# Invoice No. 4821 …",
    "usage": { "credits_used": 2 } }