Skip to main content
Back

Umi-OCR Hits 47K Stars Bringing Private Offline Text Extraction to Desktops

Umi-OCR crosses 47k stars offering a free, fully offline OCR toolkit with screenshot capture, batch jobs, PDF extraction, and an HTTP API.

Repo, Full Access
  • Umi-OCR hits ~47k stars as a free, offline, MIT-licensed OCR desktop app for Windows and Linux.
  • Covers screenshot OCR, batch jobs, PDF to searchable PDF, QR codes across 19 symbologies.
  • Layout parsing presets reorder multi-column, vertical, and code-style text automatically before output.
  • Ignore regions let you mask watermarks, headers, and footers across large batch jobs.
  • Ships CLI and HTTP API so you can wire it into pipelines as a local OCR microservice.
  • Runs on PaddleOCR-json or RapidOCR-json, installable via 7z download or Scoop.

Umi-OCR has grown to roughly 47,000 GitHub stars and 4,600 forks by packaging open-source OCR engines in a Qt/QML desktop app. The MIT-licensed project processes screenshots, images, and scanned documents on the user’s machine, avoiding cloud uploads, accounts, and usage fees. Its Windows build remains available as a portable 7z archive that can be extracted and run directly.

Local processing suits confidential documents, offline systems, and repeatable pipelines that cannot depend on a hosted API. Developers can also call the installed OCR stack through a command-line interface or local HTTP service, allowing the same engine to support desktop work, scripts, and application integrations.

From screenshots to searchable PDFs

Umi-OCR organizes its main workflows into separate tabs and bundles multiple language libraries for offline recognition:

  • Screenshot OCR: Press a global shortcut, select a screen region, or paste an image from the clipboard. Recognized text appears in a preview pane with an editable session log.
  • Batch OCR: Queue images such as jpg, png, webp, bmp, and tif, then export results as txt, jsonl, md, or csv. The app imposes no file-count ceiling and can shut down or suspend the machine after a job finishes.
  • Document recognition: Process pdf, xps, epub, mobi, fb2, and