Skip to content

Extract text from an image (OCR)

Extract the text from an image (PNG, JPEG, WebP or TIFF).

OCRimagetext-extractionPNG

Upload

Files (1–1)
One image (PNG, JPEG, WebP or TIFF), up to 20 MB. The image is sent only to these servers and is processed offline in an isolated job; it is never shared with any third-party service.
Questions people ask
What text can it actually read?
It runs Tesseract, a general-purpose OCR engine, on the pixels of your single image. It is best on clean, high-contrast, machine-printed text: documents, screenshots, and photos of text. Handwriting, low-resolution or heavily rotated/stylised text will give lower confidence; the report tells you per-line how sure it is, so you can judge. The engine is English (`eng`) only in this build.
Does the image go anywhere else?
No. The image is read in an isolated, offline job on our own servers. Tesseract is a closed local engine: it makes no network calls and there is no external OCR API behind it. The upload and the produced files are removed on the retention schedule.
What is "confidence"?
Tesseract scores each recognised word 0–100 (−1 means "not a word"). The report carries a per-line and an overall average confidence plus the bounding box for each line, so a reviewer can see which lines to re-check. A clean 95%+ average is the good shape; a low average usually means noisy pixels, not a data leak.
Why would it refuse my image?
This tool is bounded so a hostile or oversized input cannot hang it: it refuses formats outside PNG/JPEG/WebP/TIFF, images over 20 MB, or images whose decoded pixel count exceeds the 64 MP cap. Each refusal is a stable, named error (``E_FORMAT_UNSUPPORTED`` / ``E_TOO_LARGE`` / ``E_UNSUPPORTED_IMAGE``), not a timeout.
Do you keep my file?
No. The image is read, the text + report are produced, and the upload is removed as soon as that finishes. The result files stay available for about 10 minutes, then they are automatically deleted too.