A separate image version for each foreign language

The Reqser Image Translation Agent detects the images on your front end, checks which ones need a language version, and recreates those versions—in the style of the original, but with text that matches the length of the foreign-language version. The images are then automatically displayed in your store without you having to select, download, or upload a single image—right down to the filename, which is also translated for image SEO.

Reqser Image Translation Agent – Custom image versions for each foreign language

The Complete Process

Four steps that run completely automatically

From recognition in the front end to the final image version in the foreign language, the agent handles every step on its own. All you have to do is decide whether you want to intervene.

1

Play automatically

The final image versions are uploaded to your store, associated with the respective entity, and displayed on the front end of each foreign-language version in place of the original—including the translated SEO filename.

2

Recognize Front-End Images

During crawling, the agent identifies which images are actually displayed on your front end—on the home page, in banners, on category pages, and on landing pages. Images that are only stored in the media library are excluded.

3

Assess needs

Every image we find is reviewed by the AI: Does it contain text that needs to be translated into the foreign language—or is it just a product photo? Only images that actually require translation go into production. This review is free of charge.

4

Create an image version

The agent reconstructs the layout for each foreign language: the same visual style, the same fonts, the same colors—but with font sizes, line breaks, and text blocks adjusted so that the new wording fits perfectly. The wording itself comes from the same agent who also translates your product and category descriptions.

Example

One banner, three languages, three new image versions

The three images are genuine prints produced by the agent—none of them have been corrected by hand afterward. The same message requires a different amount of space in each language, so the text area is redesigned each time, while the photo, colors, font, and layout remain the same.

Front-end image Home · Promotional Banners
All languages created

verando-gartenmoebel-sale-banner.jpg

Input · Your Original

verando-meubles-de-jardin-soldes-banniere.jpg

Created by the Reqser Agent

verando-garden-furniture-sale-banner.jpg

Created by the Reqser Agent

The only input was the German image on the left—the Reqser agent generated the other two from it. The real challenge here isn’t the translation, but the length: The same sentence takes up a different amount of space in each language, and the available space is fixed. If you simply swap out the text, you’ll end up with exactly what would happen here—a French headline that extends beyond the wall and into the sofa. That’s why the agent decides for each image individually where to trim the text and where to reduce the font size, so that the language version follows the original’s graphic specifications rather than overriding them.

Everything the agent took into account in this example

  • Text length. The French headline requires about 40% more characters than the German one. This is precisely why any solution that simply replaces the text fails: The space must accommodate the new length without anything being cut off or overlapping the product photo.
  • A half-sentence doesn't make a sentence. A banner rarely contains a full sentence. “Only until Sunday” is a label, kept brief because it occupies a fixed strip of the image area. DeepL translates it as continuous text and expands it to “offre valable uniquement jusqu’à dimanche”—correct French—and the line grows from 36 to 69 characters. The Reqser Agent shortens it back to the label: numbers, the day of the week, and your dictionary terms remain; only what the translation added is omitted. However, the original text remains unchanged: “Discover now” becomes “Découvrez maintenant” and not the shorter “Découvrez”—the “now” is already in the banner; it wasn’t added by the translation.
  • Know what you're selling. For each store, we create a unique context: what’s sold there and which terms are used. The Reqser Agent reads this context before it touches an image, and it knows where the image is located—on the product page or as a banner on the front end. This is crucial for words with multiple correct translations: the same word means something different in a furniture store than in a bike shop, and what looks like it can be translated is sometimes just your product name. DeepL receives the same context for your text—so the image and the page next to it are generated from the same description.
  • The line as a whole. Each section of text is translated individually—which results in repetitions that aren’t present in the original: “Up to 40% off—only through Sunday” uses two different words, whereas French uses “jusqu’à” for both. The obvious shortened version would therefore be “Jusqu’à 40 % de réduction – jusqu’à dimanche.” The Reqser Agent reads the line exactly as it will be read later and therefore phrases the deadline differently: “dernier jour dimanche”—the same Sunday, no repetition, and the discount remains unchanged.
  • Badge and button. Neither of these are images with text on them; rather, they are areas that are sized according to the text. They are resized for each language so that the inner margin remains the same.
  • Font and Spacing. The font, style, and color remain the same. What changes are the size and spacing—just enough to ensure that the headline remains the focal point in every language.
  • Image Composition. The product photo, badge, white space, and button retain their positions and prominence. Otherwise, the language version would look like just another banner.
  • File name as an SEO signal. The name appears in the image URL and is interpreted by search engines as a ranking signal—especially in image search. A French website that returns “gartenmoebel-sale-banner” is effectively giving away this very signal. That’s why the agent translates the filename while leaving brands, numbers, and separators untouched: “verando-gartenmoebel-sale-banner” becomes “verando-mobilier-de-jardin-soldes-banniere.”
  • Format. The width and height remain exactly the same as in the original. Even a deviation of just a few pixels would cause your template to wrap differently at this point.

What Sets This Agent Apart

Deeply integrated into the store – 100% automated

Just like with your text content, you only need to maintain the default language. The crawler finds the images on your front end, identifies which ones need their own language version, creates them, and displays them. If you replace a banner months later or add a new language, the crawler will detect this during its next crawl, and the missing image versions will be generated automatically—without you having to select, download, or upload any images.

Connected to your translation agency

The text in the image goes through the same pipeline as the rest of your store's content—using your dictionary, your style guidelines, and the vocabulary that has already been translated.

Ongoing, not a one-time event

New banner online, new language enabled, original replaced—the agent detects this during the next crawl and automatically generates the missing image versions.

You remain in control

The original and the translated version appear side by side in the editor. You can also approve or reject each image if the AI review reaches a different conclusion than you do.

Detection and analysis are free

We’ll handle the entire process of capturing your front-end images and using AI to determine whether an image even needs its own language version—no matter how many images your store has.

You will be billed for each image version created

One unit is an image in a foreign language, and the analysis will determine how many of these you have. As long as an image remains unchanged, no additional costs will be incurred.

Try up to 10 image versions for free

You can test up to 10 image versions on your actual store images—within the context of your own store, using your languages and your dictionary.

This agent uses AI models trained specifically for image generation by

Google Gemini Logo DeepL Logo

Available for

Shopware Logo

Start with image translation

Connect your Shopware online store—the agent will automatically detect and check your front-end images for free. Only then will you decide whether to create the language versions.