A phone camera is the quickest scanner most people own. Photo to text is for anything printed or written in the physical world that you’d rather not retype:
- A page from a book or a printed handout you want to quote or keep notes on
- The whiteboard at the end of a meeting or lecture
- Signs, notices, posters and restaurant menus, especially when you’re traveling
- Product labels, ingredient lists, and the model or serial number on an appliance
- Letters, forms and printed instructions
Taking a photo that reads well
Most mistakes in photo OCR come from the photo, not the reading. A few seconds spent on the picture saves minutes of corrections.
- Get plenty of even light. Daylight from a window works well. Watch for the shadow of your own hand and phone falling across the page, which is the most common problem.
- Turn the flash off for glossy paper. Flash on a magazine, a laminated sign or a photo print leaves a bright patch where words simply vanish. Tilt the phone slightly instead.
- Move closer rather than zooming. Digital zoom only enlarges pixels. Stepping closer gives the text more real detail.
- Hold steady and tap the text to focus. Blur is the one problem that can’t be fixed afterwards.
- Shoot square-on when you can. Hold the phone parallel to the page, with the text filling most of the frame.
Books need extra care. Lines curve near the spine, so press the book as flat as you can and photograph one page at a time rather than a two-page spread. Labels on bottles and cans wrap around a curve, so photograph the part facing you, turn the item, and take another picture for the rest.
Fixing the photo after you take it
Auto improve is on by default. It evens out lighting and shadows, straightens a slight tilt, boosts contrast and enlarges small text, then lists the steps it took. The photo is shown as you took it; switch to the cleaned-up view to see exactly what was read.
For pictures taken at an angle, such as a whiteboard from your seat, a sign above a doorway or a page lying on a desk, use straighten with four corners. Drag each corner onto a corner of the page or board and the photo is flattened into a straight-on rectangle, so lines of text run level again. This makes a big difference to accuracy on angled shots.
You can also crop to one section, such as a single notice on a crowded board, and adjust brightness, contrast or lighting by hand if a photo is dim or has a strong shadow across it.
iPhone photos and HEIC
iPhones save photos in HEIC format by default. Image to Text App reads HEIC and HEIF directly, so you don’t need to change camera settings or convert anything. JPGs from Android phones and from iPhones set to Most Compatible work too.
Whiteboards, signs and more than one language
Whiteboards bring their own problems: faded markers, reflections from overhead lights, and handwriting. Stand where the glare disappears, even if that means an angle, and straighten the photo afterwards. If the board is mostly handwritten, switch to Handwriting mode and read the honest advice on the handwriting to text page.
Signs and labels are often printed in more than one language. You can choose several languages at once, such as English and Hindi, and Image to Text App reads both on the same image.
From photos to a document
On this page, results download as a Word file by default. Document mode joins lines into paragraphs and keeps headings and bulleted or numbered lists. Rich copy also keeps that structure when you paste into Google Docs or Word.
For something longer, such as a chapter of a book or a multi-page handout, photograph each page in turn. Every photo becomes a page you can drag into order, and Whole document view combines them into one file, with search across all of them. For more on getting an editable document out of images, see image to Word.
Limits worth knowing
Clear printed text in a sharp, well-lit photo is usually read very accurately. Expect more corrections with low light, motion blur, text photographed from far away, decorative lettering on signs, and text on crumpled, curved or shiny surfaces. Words the engine was unsure about are underlined, and the overall confidence level (high, fair or low) tells you how closely to check. For a deeper look at what affects results, see how to get accurate OCR results.