Batch OCR

When you have more than a handful of images, doing them one at a time is a chore. Batch OCR takes the whole pile at once — screenshots, scanned receipts, photographed pages, exported chat images — and turns each into text, streaming the results as it goes so you are never staring at a frozen “processing” screen.

The batch is fault-tolerant by design. A corrupt file or an image with no readable text is flagged and skipped without derailing everything else, so a 40-image job does not fail because of one bad file. When it finishes, a single combined export stitches all the text together with filenames as headers, ready to paste into a spreadsheet or archive.

Files run through the same on-device PP-OCRv6 engine as every other SafeOCR tool, one at a time rather than in parallel. That is a deliberate trade: a single WebAssembly worker moving through the queue keeps peak memory roughly constant whether the batch holds 5 images or 200, so a phone can work through a folder of forty photos without the tab running out of memory. The cost is that total time scales with the pile — recognizing a hundred receipts takes about ten times as long as recognizing ten of them, not the same time split across ten workers.

A typical batch might be a semester of expense-report receipts, a stack of scanned intake forms, or a folder of screenshots pulled from a chat thread. Because the combined export labels every section with its source filename, you can paste fifty results into a spreadsheet and still trace each line back to the photo it came from. A blurry receipt or a file that will not decode gets flagged in its own row instead of stopping the run, so you know exactly which one or two files to reshoot once everything else is already done.

Every file in the batch stays exactly where you found it — nothing is queued on a remote server, and there is no daily cap on how many you run through. Working from a single picture instead? Image to Text covers that; a multi-page document is a better fit for PDF to Text, which streams results page by page; and if one page needs to become an editable Word file rather than a plain-text export, Image to Word handles that instead.