Which reader is more accurate?

Browser reading is fast and close to perfect on screenshots and flat scans of printed text. The AI reader is better on photographs taken at an angle, creased receipts, faint print, handwriting, columns and tables.

Ten images per batch and one hundred images per day, resetting at midnight UTC. No account and no payment are needed.

Home/Tools/Image to Text

Image to Text

Copy the words out of a photo, screenshot, receipt or scan. The first pass runs inside your browser, so ordinary images never leave the device. Only when that read comes out weak is the image handed to an AI reader, and every result says which one produced it.

Your images?JPG, PNG, WebP, GIF and BMP. Ten images per batch, a hundred a day. Reading happens in your browser unless you pick an AI reader.

Drop files in, paste from the clipboard, or point at a link.

Drop images here or click to browse — JPG, PNG, WebP, GIF, BMP

Paste a screenshot Press Ctrl V anywhere on this page, or use the button below.

Link straight to the image file, not to the page around it. Other sites block browsers from reading their images, so this one hop goes through our server and nothing is kept.

Working...

Images 0

Reading

Applies to every image in the batch.

Reader

Language in the image

Today

0 of 100 images 0/10

Browser reading keeps the image on this device. The recogniser downloads once, then works offline.

How to use it

Three steps, and only the middle one has a choice in it.

Add the images

Drop up to ten files in, paste a screenshot straight from the clipboard, or paste a link to an image hosted elsewhere. Photos, screenshots, receipts, book pages and scanned documents all work.

Leave it on Auto

Auto reads on your device first and only sends an image on when that read scores badly, which is what happens with handwriting, curved pages and low light. Force it either way with the two other settings: browser only never uploads, AI reader always does.

Fix and copy

Every result lands in an editable box. Correct the odd character, then copy one block, copy the lot, or download everything as a text file.

Questions

Are my images uploaded?

Only when the reader on your device cannot cope. On the Auto setting every image is read locally first, by an engine downloaded to your browser, and it is sent on only if that read scores below 90 percent confidence or finds nothing. Set the reader to browser only and nothing is ever uploaded, at the cost of accuracy on hard pictures. When an image does go to the AI reader it is sent to a model provider for the length of the request, is not stored by us, and is not used to train anything. Adding an image by link is a separate case, where our server fetches the file once and passes the bytes straight to your browser.

Which one is more accurate?

For screenshots, PDFs turned into images, and flat scans of printed text, browser reading is fast and close to perfect. The AI reader wins on anything awkward: photographs taken at an angle, creased receipts, faint thermal print, handwriting, columns, and tables that need their shape preserved. The tool measures its own confidence after a browser pass and offers to re-read weak results for you.

What are the limits?

Ten images in one batch and a hundred reads a day, resetting at midnight UTC. A read is one pass, so an image that Auto sends on to the AI reader uses two. Nothing is charged and no account is needed. If every free AI model is busy, which happens at peak times, the page offers a paid model instead. It stays free for you; the cost is ours.

Can it read handwriting?

The AI reader often can, especially neat handwriting on a plain background. Browser reading essentially cannot: it is trained on printed type and will produce noise from a handwritten page. Pick the AI reader before you start rather than after.

How do I get a better result?

Fill the frame with the text, get the page square to the camera, and use even light without a shadow across the middle. Sharpness matters more than resolution, and a photo of a screen will always read worse than a screenshot of it. If a language other than English is in the picture, set it in the language box first, because the wrong setting is the single most common cause of garbled output.