AI Image Translator for Photos and Screenshots
Drop a PNG, JPG, or WebP file with printed words. Optical character recognition finds the lines, the source language is detected for you, and the editing model redraws those lines in the language you pick — then you compare and download the preview without leaving this URL.
Upload an image or drag & drop
PNG · JPG · WebP · max 10 MB
Try a sample
English-text samples first — menus, signs, screenshots, products & comics
Source language is detected automatically
Upload an image and click translate. The translated preview and extracted text appear here.
Real translation example
See a restaurant menu translated in place
An English curry menu redrawn into Japanese while food photos, prices, and the original layout stay in view — the same class of job as a sign or a screenshot.
Restaurant menu · English → Japanese
Curry dishes, prices, descriptions and reservation footer translated and redrawn on the original menu design.
Drag to compare
Generated on this site at 2K (6 credits). Upload your own menu, sign, screenshot, or product label in the form above to try the same workflow.
How this AI image translator works
Upload a readable still
Choose a sharp menu photo, street sign, app screenshot, packaging label, or product picture. Printed type with contrast beats handwriting, motion blur, and tiny captions. PNG, JPG, and WebP up to 10 MB work; sample files on the form let you test the pipeline before using your own shot.
Let OCR read the visible words
You do not crop boxes or retype. Vision OCR locates printed strings on the still, groups them into regions, and guesses the original language. Crooked posters and busy restaurant layouts still work better when the type is large enough to read on a phone.
Pick a target language and review
Choose English, Spanish, Chinese, Japanese, Korean, French, German, or Portuguese. Each run costs 6 credits and may take up to two minutes. Drag the compare slider, skim the region list, then download a PNG preview. The original file on your device is not overwritten.
What you get after one upload
OCR on photos, not just documents
The reader is built for camera stills and screen captures: menus, notices, UI chrome, labels, and packaging. You skip a separate scanner app and a second paste into a text-only translator.
Source language without a guess
You only choose where the wording should go. The model infers the original language from the pixels, which helps when a sign mixes scripts or you cannot name the language you are looking at.
Eight destinations on one form
English is the default. Spanish, Chinese, Japanese, Korean, French, German, and Portuguese cover travel, study, and marketplace research without hopping between single-language gadgets.
Before-and-after slider
Judge whether dish names, prices, and legal lines landed in the right blocks. The slider sits on the workbench and again on the restaurant-menu example below so you can see a real redraw before you spend credits.
Region list for spot-checks
Each detected block shows original and translated wording side by side. Use it when a price, allergen line, or button label matters more than a pretty overlay.
PNG preview you can save
Download the redrawn still for a trip, a class, or a product teardown. Treat it as a reading aid: complex comics and dense contracts may still need a human check.
When photo translation helps — and when to use something else
An AI image translator is for wording that lives in pixels: a photographed menu, a platform screenshot, a shipping label, a museum plaque. It is not a typesetting studio and not a substitute for a certified human translator on contracts or medical leaflets. Upload, detect, redraw, compare, download — the landing URL and the product are the same view.
How in-place redraw differs from copy-paste OCR
Classic workflows dump recognized characters into a notes app and you lose which line sits on the burger photo. Here the editing model tries to paint the new wording back onto the still so you can read the scene as a scene. Dense manga, neon night shots, and seven-column wine lists can still show small glitches — the job is visual, not a Word file. If you only need a transcript, copy the region list; if you need to understand the picture, use the redraw.
What a good source file looks like
Hold the camera parallel to the page and fill the frame with the type. Avoid glare, extreme wide angle, and photos taken while walking. Screenshots should be the full window, not a photo of a monitor. Handwriting, embossed metal, and tiny captions are the usual misses. If a first pass drops a line, recrop tighter or shoot again in daylight.
Travel menus, signs, and packaging
Use this when a restaurant has no English sheet, a station notice is only in the local script, or a box lists ingredients you cannot parse. Photograph one spread at a time and check footer regions for prices and set-menu rules. Shoot the panel you care about (allergens, size chart, how-to icons) instead of the whole shelf. The showcase below is an English curry menu redrawn into Japanese with layout and food photos kept.
Screenshots, games, and product research
A screenshot translator is the same form with a different file: an app dialog, a game quest, a marketplace size chart, or a checkout you cannot switch. Avoid password managers and private chats. Full-resolution captures beat zoomed thumbnails. After you understand the still, related Pixwit tools can caption or convert the file if you need another format.
Credits, wait time, privacy, and limits
Sign in, confirm the 6-credit cost, and wait up to two minutes while the model reads and redraws. Failed jobs should be retried; the original upload stays on your device until you reset. Do not send passports, medical records, or other people’s IDs. Faint watermarks, stamp numbers, and type that matches the background can be skipped. Only the eight languages on the selector are available. If one badge is wrong, trust the region list and treat the pixels as illustration.
Background reading: Optical character recognition (Wikipedia)
Where an AI image translator gets used
Restaurant and cafe menus abroad
Photograph the page in front of you, run it toward your language, and read dishes, extras, and prices without typing every line. The slider is there so you can match a photo of the plate to the translated name before you order.
Street signs, tickets, and hotel notices
Platform changes, checkout times, and safety stickers are often posted as graphics. A clear still is faster than finding a staff member who can paraphrase, especially late at night in a station you do not know.
App, game, and website screenshots
When a setting panel or quest text has no locale toggle, capture the screen and translate the chrome. Keep captures free of account secrets. This is a reading aid for one screen, not a patch for the whole product.
Packaging, manuals, and marketplace photos
Ingredient panels, size charts, and how-to diagrams on listings are frequently images. Translating the still helps you compare products, spot warnings, and decide whether a manual page is the one you need.
Study, comics, and language practice
Learners photograph worksheets, posters, and comic panels to see a gloss next to the original art. Use the region list to study phrasing. Stylized lettering and overlapping balloons may need a second, tighter crop.
Team research and localization previews
Researchers and designers drop competitor screenshots or packaging shots into the form to brief a writer. Treat the PNG as a conversation piece; ship-ready copy still goes through your localization process.
Questions about translating image text
Related tools
Translate the next still on this page
Use the upload box above for a menu, sign, screenshot, or label. Each run costs 6 credits and returns a redrawn preview you can download.