Load a photo or screenshot with printed English text, read the text out of it on your device, and translate it - the photo never leaves your browser.
Example - what a result looks like (no download needed to view this):
Text the on-device reader produced: Your photo never leaves the browser The text is read and translated here Menu prices total 45 dollars today
Spanish translation the on-device model produced: Su foto nunca sale del navegador El texto se lee y se traduce aquí Menú precios totales 45 dólares hoy
Both outputs above were produced ahead of time by the same reader and model this page runs, so you can see the output shape before loading anything.
Photo Translator - Translate Text in a Photo, On-Device
Point this page at a photo, a screenshot, or a scan with printed English text - a menu, a sign, a package label, a slide, an error dialog - and it reads the text out of the image and translates it, entirely in your browser. The photo is never uploaded: the reading runs on your device with an engine served from this site, and the translation runs on your device too, either with Chrome's built-in translation or with a small model you download once and your browser keeps.
The flow is two visible steps. Step 1 reads the text: pick the image, click the button, and the recognized text appears in an editable box - so when the reader trips over a blurry word you can correct it before translating, instead of translating the mistake. Step 2 translates that text into your pick of Spanish, French, German, Italian, Portuguese, Indonesian, Vietnamese, Hindi, Chinese, Russian or Arabic. If you skip Step 2 entirely, Step 1 is still a complete result: copyable text pulled out of an image with nothing downloaded beyond the small reader.
Two boundaries are worth knowing before you rely on it. The reader is built for clear printed text and reads English; handwriting belongs on the handwriting to text page, which runs a model trained for it. And the translation quality is small-model: solid for everyday sentences, weaker on idioms and names, so review anything important before you act on it.
If you already have the text as text, skip the photo step and use the AI translator directly - it adds auto language detection and more direction choices. If you want the text out of an image with no translation at all, the image to text OCR page does exactly that one job.
Frequently Asked Questions
Is my photo uploaded anywhere?
No. The text is read out of the photo on your device by an engine served from this site, and the translation also runs on your device - with Chrome's built-in engine or with a model your browser downloads once and keeps. There is no server that ever sees the image or the text.
What does it download, and when?
The reader fetches about 5 MB from this site the first time you click "Read text from photo". Translation needs either nothing (when Chrome's built-in translation covers the pair) or a one-time model of about 115 MB per language pair, and the size is shown on the button before you start - nothing downloads without your click.
Which languages does it translate to?
The text in the photo is read as English, and you can translate it into Spanish, French, German, Italian, Portuguese, Indonesian, Vietnamese, Hindi, Chinese, Russian or Arabic. Pairs that the on-device model cannot serve are not listed, so every option in the menu actually works.
Why did it misread part of my photo?
The reader works best on sharp, well-lit, straight-on printed text; blur, glare, tiny fonts and steep angles cause slips. The recognized text sits in an editable box exactly for this - fix the slip there, then translate. Handwriting is a different problem and has its own page with a model trained for it.
Does it replace the text inside the image?
No. The result is text - the recognized original and its translation, both copyable. The photo itself is not re-rendered with translated text drawn over it.