You have a picture of some words — a photo of a page, a screenshot, a receipt, a sign, a slide — and you want the words as text you can edit, copy or search. Retyping is the obvious way, and for a sentence or two it is the fastest. For anything longer, the job belongs to OCR.
OCR — optical character recognition — reads the pixels of an image and works out which letters they represent. It finds the regions that look like text, breaks them into lines and characters, matches each shape against a trained model, and outputs its best guess as real, editable text.
The key word is guess. OCR is not revealing hidden text; there is no text in a photo, only shapes. It infers characters from those shapes, which means it can be wrong — and it is wrong more often when the image is small, blurry, skewed or badly lit. Get the picture right and modern OCR is remarkably accurate. Feed it a dim, crooked snapshot and no engine will save you.
| Helps a lot | Hurts |
|---|---|
| Sharp focus, text filling the frame | Small or distant text, motion blur |
| Straight, level lines of text | A tilted or curved page |
| Even lighting, dark text on light background | Glare, shadow across the page, low contrast |
| Ordinary printed fonts | Handwriting, decorative or tiny type |
Not to the OCR itself; they are all just images. But screenshots have an advantage: they are pin-sharp and evenly lit by definition, so text from a screenshot usually recognises almost perfectly. A phone photo is at the mercy of your lighting and steadiness — which is why, for a full page, a deliberate scan beats a casual snap. See scanning documents with a phone for how to capture a clean page.
Pennary's text recognition works on the page in front of it, on your device.
.txt file or exporting a searchable PDF is a Pro feature.Even a clean image produces the odd mistake, and OCR errors cluster in predictable spots. Check these before you trust the text:
Most online OCR sends your image to a server to be processed. The things people most often photograph for text — receipts, letters, ID cards, medical or official documents — are exactly the things you would rather not upload. Pennary recognises the page locally in your browser, so the image never leaves your machine. It is a little slower than a data centre; the privacy is the point.