hiroi-sora/Umi-OCR ? reverse-engineered prompt

Build me a free offline OCR desktop app like Umi OCR that works on Windows and Linux. I want a simple app where I can grab text from screenshots, paste or drag in images, process lots of images at once, and handle PDF or scanned documents too. It should let me copy the recognized text, save results to common file formats, and clean up the output so paragraphs and multi column layouts read naturally. Please also include a way to ignore parts of an image like watermarks or headers and footers, plus support for reading and generating QR codes and barcodes. The app should run without needing internet, support multiple languages, and feel polished with a clear settings page and a few themes. If anything is unclear, look up the current docs online and then build the best working version you can.