Translate text in image

Use Translate for a product graphic, poster or app screen you are allowed to localize. Choose a target language after uploading; the model rewrites visible text while trying to retain the image's layout and lettering style.

Translations can be longer than the original and may not fit a small label. Review every line with a fluent speaker, particularly names, offers, numbers and right-to-left text.

A clean, high-resolution source makes small labels easier to read. If the artwork contains dense paragraphs or multiple scripts, inspect the output closely and consider editing shorter sections separately.

For a production design with an editable source file, use that file when you need exact line breaks or brand typography. The image editor accepts one JPG, PNG or WebP at a time.

Translation mode asks the model to translate all visible text and redraw it in the image. Unlike copying text into a separate document, the intended result is a new image. The source artwork, names and any wording that should remain unchanged still need careful review.

Choose one language per edit. A second language requires another submission and another credit. Translations can vary, especially for uncommon wording, small labels and mixed scripts.

Ask a fluent speaker to check meaning, punctuation, proper names, numbers and cultural context. Long translations can exceed the space available, and right-to-left scripts can expose alignment problems.

The model is instructed to retain framing and aspect ratio, but the output dimensions may change. For a specific print size, app-store requirement or exact typography, check the file and use the editable source design when necessary.

A reading aid can extract or overlay translated text so you understand a sign or screenshot. An image translation workflow creates a new image with visible wording redrawn in the selected language. Choose according to the job: reading a message is different from preparing a localized graphic for review.

TextImageEdit's Translate mode asks the model to translate visible text in one uploaded image and keep the surrounding layout and style. The text examples on this page illustrate possible wording; they are not claimed to be actual image-translation outputs.

Upload one JPG, PNG or WebP up to 10 MB, choose Translate and select the target language. The available choices include English, Spanish, Portuguese, French, German, Italian, Japanese, Korean, simplified and traditional Chinese, Arabic and Russian.

The editor creates one target-language version per submission. To prepare several languages, run and review a separate image for each one. Each submitted image costs one credit; the first Google account credit is free.

A fluent speaker should check the translation before it is published. Product names, abbreviations, dates, measurement units, discounts and idioms can be mistranslated or reformatted. A number that appears correct visually can still communicate a different offer in another locale.

Where a phrase is legally required or safety-critical, obtain the approved wording from the responsible organization and use the editable source design. An automatically generated translation is a draft to review, not a substitute for professional approval.

Translations do not always occupy the same width as the original. A longer word can overflow a button or force an unexpected line break. Languages with different writing direction or characters may need different spacing; check the whole image as well as the edited line.

The model is asked to retain the framing and aspect ratio, but it may change fine details or output dimensions. Verify the generated image against the original and against the publishing destination's size requirements.

For an app screenshot, the localized text must match what users can actually see in the product. For a menu, review dish names and prices. For a campaign graphic, make sure the localized offer is valid in the target region and that the text is readable at the final display size.

A design source is the better choice when layouts vary by language or pixel dimensions must be controlled precisely. A finished image is convenient when the source is unavailable, provided the generated version is reviewed and used within the rights granted for the original.

Translate mode is useful when you want the model to translate visible text into one selected language. Replace text is a different workflow: you provide the source wording and your own approved replacement. Choose the mode that matches whether you need the model to translate or already have the final wording.

For a campaign with legally reviewed copy or a product label with approved terminology, use the approved phrase and review it in context. Automated translation should be checked by a fluent speaker before it is treated as final.

A poster or product graphic may contain a headline, subheading and several labels. Translate mode asks the model to translate visible text throughout the image. Review each area, including fine print and numbers, because an image model can miss small characters or change unrelated details.

If only one line needs a change and the exact new words are known, Replace text gives a more explicit old-to-new instruction. One submission works on one image; additional language versions require separate submissions.

A phrase can become longer or shorter when translated. A button, badge or narrow column may not have room for the new wording, and a longer translation can wrap into adjacent artwork. Check spacing, line breaks, alignment and whether the text is still legible.

Some scripts read in a different direction or need different letter spacing. Ask a fluent reviewer to inspect the actual output, not just the words in a separate document. If the layout needs substantial adaptation, use the editable design source.

For an app store, verify that the screenshot represents the localized product and meets the store's current dimensions. For a menu or price list, verify numbers, currency and dietary information. For a product graphic, confirm that the translated claim and offer apply to the destination market.

The generated file can have different pixel dimensions or format from the upload. Check the downloaded asset before publishing or printing it. An illustrative translation shown on this page is a wording example, not a claim that a specific image was generated.

Start with an owned or authorized graphic and choose the target language. Translate mode asks the model to translate visible text in the image. The illustrative wording on this page shows how a short English headline might be expressed in Spanish; it is an example of wording, not a claimed generated image.

Ask a fluent speaker to review the final text, numbers, units and product names. Confirm that any localized offer is valid in the destination market, and update the store listing separately. The editor returns an image; it does not publish or change a product page.

A restaurant may want an additional-language version of its own menu graphic. Upload a clear supported image and select one target language. Small dish descriptions, prices and dietary labels need especially careful review because the model can miss or alter fine characters.

A fluent reviewer should check dish names, allergens, currency and the context in which the menu will be used. If the business maintains a layered source or a menu database, update that source as well; an edited image is not an update to the underlying menu.

A screenshot's UI labels and marketing copy should reflect the language available in the product. Translate one target-language image per submission, then compare the result with the actual localized application. Review text expansion, icons, buttons and status details.

A generated image may not preserve the upload's exact pixel dimensions. Before submitting to an app store, check the actual file against the current platform requirements. Where possible, a fresh capture of the localized app gives reviewers a direct view of its real behavior.

Translate mode is designed to request an automatic language version of visible text. If you already have a specific approved phrase, Replace text allows you to enter the old and new wording directly. Choose that mode for a single headline when exact wording matters more than translating every visible line.

A legal, safety or medical statement may require specialist approval and controlled source artwork. The image model does not replace a translator, reviewer or source design. Keep the original file and check the output against the approved content.

A camera or text-recognition translator may be enough when the immediate goal is to understand a sign, label or screenshot. If the goal is to prepare an image in another language, the workflow includes different checks: the translation must fit its visual space, the result must be downloaded, and the words must be correct for the intended audience.

TextImageEdit produces an edited image after you choose a target language. It does not provide an editable translation table or update text in the source application. Keep the original graphic and have a fluent reviewer compare both the meaning and the resulting layout before the translated image is published.

A campaign image may contain a headline, a subheading and a discount. Translate mode requests a new-language version of visible text in the image. A short, clearly separated headline is straightforward to inspect; a dense infographic with many small labels needs a more careful line-by-line review.

When only one phrase should change and a human has already approved its exact wording, switch to Replace text and enter that original and replacement. This distinction matters for proper names, legal copy and brand slogans: a model-generated translation may not be the organization's chosen phrasing.

The same idea can occupy different amounts of space in different languages. A translated button label may be longer than its source, or a headline may wrap where the original did not. Check that words do not overlap borders, cover illustrations or become too small to read.

Right-to-left text and scripts with connected characters also deserve a fluent review. Inspect the order of words, punctuation and numerals in the actual image, not only in a separate translation draft. If the entire composition needs to be reflowed, use the editable source design.

Make a list of the source image, target language, approved terminology and intended destination before starting. Choose one language per submission and save each reviewed result with a clear filename. Separate files help avoid uploading the wrong language to a market, store listing or help article.

Review every version independently. Check that dates, prices, currencies and units are appropriate for that market, and verify file dimensions against the destination's current requirements. Each image submission uses one credit; the first Google signup credit can be used for a first translation.

The wording examples on this page demonstrate possible source and target phrases for a product promotion, a menu and an app screen. They are illustrative text, not screenshots of an automatically generated translation. The actual model output can differ and requires review.

For a real product screenshot, confirm the translated interface is actually available to users and that the image represents it accurately. For a menu or product label, check that names, prices and any safety information remain correct. The source image must be yours or authorized for editing.