Umi-OCR Hits 47k Stars for Local Desktop Text Extraction

The open-source optical character recognition application Umi-OCR has amassed roughly 47,000 GitHub stars and 4,600 forks, emerging as a popular offline tool for desktop text extraction. Distributed under the MIT license for Windows and Linux, the Qt/QML desktop application runs entirely on the user's local hardware without requiring cloud uploads, user accounts, or subscription fees. The software can be deployed as a portable 7z archive or installed via Scoop, leveraging underlying recognition backends such as PaddleOCR-json or RapidOCR-json.

Designed for both individual tasks and high-volume automation, Umi-OCR handles screenshots, images, and multi-page document conversions. It supports standard image formats including jpg, png, webp, bmp, and tif, as well as complex document formats like pdf, xps, epub, mobi, and fb2, enabling users to generate searchable PDFs. The tool also extracts QR codes across 19 symbologies. Advanced layout parsing presets automatically reorder multi-column, vertical, and code-style text before producing exports in txt, jsonl, md, or csv formats. Furthermore, users can set ignore regions to filter out repetitive watermarks, headers, and footers across batch jobs that carry no file-count limits and can trigger machine shutdown or suspension upon completion.

For practitioners and developers, Umi-OCR bridges the gap between interactive desktop utilities and automated data pipelines. Beyond its graphical interface—which includes features like global screenshot hotkeys and editable preview logs—the app embeds a command-line interface and a local HTTP API. This allows teams handling confidential documents or working in air-gapped environments to deploy the tool as a local OCR microservice, bringing privacy-focused, scriptable document processing into existing backend workflows.