Canva image translator: how it works and how it compares

Canva can translate the text in an image, but not in one step. Its image translator is a workflow inside the editor: Grab Text turns the words in a photo into editable text boxes, then the Translate app translates them. That works for straight, Latin-script text, and you end up with a design you can keep editing. TranslateMyImage takes a different route: an AI image model regenerates the whole image with the translated text in the original layout. Which one fits depends on the image and on what you will do with it next.

What Canva does with the text in an image

The documented route: Grab Text, then Translate

Canva's image translator page presents the feature as part of Canva Pro and describes these steps:

1. Upload the image through the Uploads tab, into a new design or a template.
2. Select the image, choose Edit, and under Magic Studio pick Grab Text. Select all the text, or only the parts you want.
3. Open Apps and search for Translate. Pick the target language and, optionally, a tone, then translate the page or the selected text.
4. Adjust the design, then download it as JPG, PNG or PDF.

Grab Text is what makes this possible: Canva's help center describes it as turning text that appears inside photos and images into editable text boxes (Grab Text help). The same article lists its limits. It works with Latin text only; it does not support rotated or arched text, languages that do not use the Latin alphabet, or files with more than one page; and it does not process multiple images or pages with a single click. Canva also notes that Grab Text tries to match the font and position, may leave small artifacts of the old text (it suggests Magic Eraser for those), and can occasionally create text from patterns or line drawings. Canva's image to text page says Grab Text comes with a Pro subscription.

The Translate app itself

Translate is Canva's machine translation tool for the editable text in a design. As of October 2026, Canva's Translate page says it supports 134 languages (language list), works on all design types, keeps the design when translating between left-to-right and right-to-left languages, and can be tried once for free before a Pro, Teams, Nonprofit or Education subscription. Canva's help center adds that you translate several pages into one language at a time, within a 2,000-character limit, that Translate does not work with every font, and that it can also be run from Resize under Magic Switch. By default, Translate reduces the font size to fit and adds the translation on duplicate pages.

On usage, Canva's AI usage help says Translate does not count toward the monthly AI allowance on paid plans other than Pro Lite, with fair-use limits; on Canva Free and Pro Lite these tools draw from a monthly allowance.

Magic Layers: flat graphics into editable layers

Canva also has Magic Layers, which separates a flat PNG or JPG into editable layers, including text you can edit directly (Magic Layers, marked Beta). Its help article says it works best with graphic designs, illustrations and stylized visuals, that photo-realistic images may not separate cleanly, and that it does not take PDFs, multi-page files or images over 5,000 px on the longest edge. It is a premium AI tool, so each use counts toward the monthly AI allowance. Canva does not present Magic Layers as a translation feature, and its pages do not say which scripts it recognizes, so we make no claim about using it to translate.

What TranslateMyImage does differently

TranslateMyImage sends the image to a generative AI image model that reads the text, translates it and redraws the whole image in a single pass. There is no separate step to extract the text, erase it and retype it, so the source text does not need to be Latin-script, straight or editable first. The model is instructed to keep the layout, typography, colors and non-text elements unchanged.

• Source languages: auto-detected, or chosen in the uploader, including Japanese, Korean and Chinese.
• Modes: Default, Marketing (persuasive wording, calls to action), Manga/Comics (speech bubbles, sound effects) and Technical Documentation (annotations, precise terms, measurements).
• Batch translation from the Pro plan: up to 10 images per run, each in one or more languages, up to 50 outputs, downloaded as a ZIP.
• An API, described in the API documentation, from the Pro plan.
• Price: the free Basic plan gives 3 credits a month for personal use, and 1 credit translates 1 image; Pro costs $9.90 a month for 100 credits, Premium $19.90 for 300. See pricing.

The trade-offs are real. The result is an image, not text boxes: to change a word, you translate again or retouch the image. Because the image is regenerated, digits, small print and logos can change, and the pixel size can differ from the original, so every result needs a check against the source. The uploader translates into 9 languages (English, French, Spanish, German, Italian, Portuguese, Japanese, Korean, Chinese), and the language pages cover 45 target languages, far fewer than Canva's 134. There is no design editor, no PDF input and no camera mode.

Feature comparison

Choose Canva when

• You are building or editing the design in Canva anyway. Translate works on the editable text of the whole design and, by default, adds the translation on duplicate pages.
• You will keep changing the wording. Text from Grab Text stays editable: you can correct a term, restyle a headline or resize the design later.
• The text in the image is simple, straight and in a Latin-script language. A short slogan on a plain background is exactly what Grab Text is built for.
• You need a language our tool does not cover. Canva lists 134 target languages.
• Your team works with professional translators. Canva's help center describes a Smartling app that sends design text to a Smartling translation project and generates the translated versions in Canva.

Choose TranslateMyImage when

• The source text is not Latin-script. Japanese packaging, a Korean poster or a Chinese screenshot: Canva's help says Grab Text does not support these.
• The text is styled, curved or part of the artwork. Grab Text does not support rotated or arched text; the image model is instructed to keep the lettering where it was.
• The image is the deliverable and you do not use Canva. Upload, translate, download, without building a design.
• You have many images or many languages. Batch runs and the API are built for that.

Not sure? Test both on two or three of your real images. For other tools, see the DeepL comparison and the Google Translate comparison.

Using both on the same design

The two tools fit together well on a Canva project. Run Translate on the design's editable text, which covers most of it. For the photos and uploaded graphics that Grab Text cannot handle (non-Latin text, curved lettering, busy backgrounds), download them, translate them with the image translator and put them back with Replace. Our guide on how to translate a Canva design walks through that workflow and the checks to run before exporting.

Canva is also a convenient place to fix a generated image. If one word in a translated image is wrong, cover it in Canva and set the correct text on top, rather than translating the whole image again.