Your listing is live in a new Amazon store, and the copy reads like it belongs there. The first photo still speaks the language of your home market, though, and shoppers can see it before they reach a single bullet. The instinct to translate Amazon main image text makes sense when that photo carries so much of the first impression.
The problem gets harder as your catalog crosses more borders. A polished listing can feel finished inside Seller Central while presenting a different message on the customer-facing page. Every new store makes it uncomfortable to see how much of the experience you have actually checked as a local shopper would see it.
Cross-border copy cannot fix a text-heavy first photo
You push a listing through Build International Listings on a Tuesday morning, and by afternoon the German store shows a title, bullets, and specifications in German. Machine-generated, connected to your source store, and live without a single manual edit. Then you open the German detail page and scroll the images.
Amazon’s guidance already tells sellers to translate the text inside product images and videos for the target store, and the August 2026 announcement moves in that direction by automatically translating text in eligible images across stores in 14 countries. The translated versions appear on detail pages and in Image Manager, your original stays untouched in the store’s primary language, and uploading your own correctly translated image removes Amazon’s version. That covers the gallery, and it covers it with reasonable care for the seller’s control.
The first photo sits outside that arrangement. Amazon said the main image would not be translated, and its main-image rules already prohibit text and other graphics on it. The first photo is the one that appears in search results, so it is the image a German shopper meets before any translated bullet exists. If your hero shot carries a badge, a unit callout, or a claim in English, it will be read as English, and because the rules forbid text there, the cleaner fix is often a main image that needs no language at all.
What you can rely on beyond that is narrower than it looks. Country-Specific Upload lets brand owners upload local-language images, yet Amazon said those uploads weren’t guaranteed to replace global images on the detail page. Forum reports add texture: automated translation can miss stylized fonts, logos, or packaging text, and one seller reported uploaded Canadian images not appearing when they weren’t the listing creator. Those accounts are anecdotal, and one predates the 2026 rollout, so treat them as warnings about where to check rather than proof of how the system behaves today.
Translation boundaries: Amazon translates gallery text, not main images
The boundary comes down to a difference in jobs. The main image is the first image shown in search, and Amazon’s guidance prohibits text, logos, borders, watermarks, and any added visual content in it. The additional images get the opposite brief: product features, dimensions, usage instructions, and the product shown in use or from other angles. Text belongs in one place and is forbidden in the other, and the translation feature follows that line. It works on the gallery images, where language does the work, and leaves the main image alone.
For a seller who wants to translate Amazon main image text, the first move is to check whether there is any text to translate. A main image that complies with the rules carries no words, so nothing is left behind in English for a German or Polish shopper to misread. The gap only opens when a hero shot has drifted from the guidelines, with a badge or callout baked into the photo. In that case Amazon’s translation won’t rescue it. The fix is in the file you upload.
Secondary images are where the feature does real work. Across the 14 stores covered, a shopper whose language differs from the original sees a translated version of your dimension diagram or instruction panel, while the original remains unchanged. You can still take over from Amazon by uploading your own correctly translated image to the detail page. That control matters most for the images carrying the claims a buyer weighs before purchase.
Automation here is also conditional. Complex graphics, stylized fonts, logos, and packaging text can slip past detection, so an infographic with decorative lettering is a likelier candidate for a missed line than a clean, high-contrast layout. Because detail pages can draw on several contributors, the image you uploaded is not guaranteed to be the one a shopper sees. Auditing each store’s gallery in the shopper’s language therefore stays part of the job, even where translation is switched on.
Picture what shoppers in Germany see after all this: a wordless main image, then a gallery of dimensions and instructions that read naturally in German, each matching the language of the bullets beside it.
First-image compliance makes text-free photos work across stores
A wordless first photo is the only version of it that every store can read. Amazon’s rules ask for the product alone against a white background, with no text, logos, watermarks, or other graphics layered on top. Its global-selling guidance separately tells you to translate text in product images into the language of each targeted store, and the two instructions fit together without strain: the main image has nothing to translate because it has nothing written on it.
That design pays off the moment you look at how Amazon’s automatic image translation behaves. The feature, announced across 14 stores from the US and Mexico to Poland and Sweden, renders text in product images into the customer’s preferred language and keeps the original untouched. Amazon does not translate text in the main image, so a badge or callout baked into the hero shot gets no localization from Amazon. Adding words back to rescue a weak photo also carries a cost, because Amazon says non-compliant images may be rejected, removed, or altered, and the listing can be suppressed from search results.
Language-neutral photography means letting the object carry its own message. Show the product at an angle that reveals its form and scale, keep the background clean, and avoid packaging with prominent printed claims facing the lens. Where the product itself bears text, such as a retail box or a printed label, choose the angle and crop that make the item identifiable without turning the label into the selling point. Anything a buyer needs spelled out, whether dimensions, materials, or usage, belongs in the secondary images, the bullets, or A+ Content, where translation is expected and where you can replace Amazon’s version with your own correctly localized upload.
The compliance question and the localization question turn out to share one answer. A hero image built to meet the white-background rule already works in every store, in every language, with no external translation step to fail.
Recognition failures can leave stylized image text untranslated
Everything Amazon does translate is only as good as what its system can detect. Amazon says its image-translation system applies AI with quality guardrails before a translated image goes live, after which the result appears in Image Manager and on the pages for individual products. A guardrail checks the output it was given, though, and text that was never recognized gives it nothing to check. Seller Central guidance acknowledges the gap: text in complex artwork, decorative lettering, brand emblems, or on product packages can go undetected and remain in its original language within an otherwise translated image.
This matters most in your secondary images, where the real selling happens. A lifestyle shot with a flowing script headline, a badge built from a custom typeface, or a photo of retail packaging whose printed claims curve around a bottle are exactly the cases where partial translation can leave a Spanish or German shopper looking at half-English artwork. The packaging problem is not unique to Amazon: research on reading text from product packaging names dense layouts, varied fonts, glare, and curved surfaces as persistent obstacles for automated recognition. Your design choices decide how many of those obstacles your image carries.
Seller Central’s own advice is plain: use clear, high-contrast standard fonts, and keep simple text separated from complex backgrounds. A bold sans-serif callout on a flat color block gives the system something it can isolate. The same words in a decorative face over a busy photograph give it a puzzle, and a logo wordmark will likely stay as drawn, which is usually what you want for a brand name anyway.
The fallback is already built in. The original image in each store’s primary language stays unchanged, and you can upload your own localized version to replace the automatic one. For any image where stylized type is central to the design, a translated file you made and checked yourself removes the dependence on detection entirely, and a mistranslated slide never reaches the shopper who is deciding whether to trust your listing.
Asset operations: Localized uploads may not replace global images
Uploading a localized image is only half the job, because Amazon’s own wording leaves room for it not to show. Image Manager is where you find out what shoppers in each store actually see, and it gives you more to work with than most sellers use. It splits global images from country-specific ones, and recent updates let you upload country-specific images from the start and filter by which images are displayed in particular stores. Run that filter store by store before you trust any localized file.
The logic of your edits follows from how Amazon structures the swap. Translated images appear on the product detail page and in Image Manager, while the original, untranslated image remains untouched underneath. When you upload your own version, Amazon says its generated translation comes off the page, so your file becomes the only localized option. Deleting a country-specific image is also allowed, which gives you a clean reversal when a localized file turns out to be wrong.
So a verification routine has a natural shape: after each upload, open the filtered view for the target store, confirm your file is the one marked as displayed, and compare it against the live detail page. The reason to repeat this is that Amazon says a country-specific upload may leave the page’s global images in place, and sellers have reported local-language images in EU marketplaces being overwritten by English ones, a pattern consistent with a detail page that draws images from several contributors. An upload that sits in your library while an English image holds the live slot has done nothing for the Spanish shopper, and only the check reveals that.
The routine covers secondary images, and it leaves the first photo outside its reach, since Amazon does not translate the main image through this workflow.
Investment choices favor wordless main images and tested secondary localization
Your spending logic has to start from that gap, because the first image is the one Amazon says appears in search results and opens the detail page. Translation on the secondary slots is automatic, and a replacement upload gives you control where you want it. The main image is the single position where no translation arrives, so any language problem there is yours to solve.
The workaround budget is narrow by design. Compliance rules govern the main image, and Amazon warns that non-compliant images can be rejected, removed, altered, or suppressed from search results, so a workaround that adds text or graphics to a file the rules constrain invites a penalty that costs more than a missing translation. Because each SKU is submitted in only one language through inventory templates, you may also face a choice between keeping a single main image or maintaining separate listings per language. The safest investment is usually to let the product, its scale, and its context carry the main image without any words.
Localization budgets have a clearer target in secondary images. Amazon’s automatic translation may leave some text in product images untranslated, so clear, high-contrast standard lettering and simple layouts give the system more to detect. Where a graphic is too complex to survive that process, a human-translated replacement removes Amazon’s version and puts your wording in its place. Professional translation services are the route Amazon itself points to for A+ Content, which suggests the same treatment for image text that carries your selling argument.
What you will not find is a ranking of these options by return. Amazon’s translation guidance offers no seller-adoption percentage and no conversion-lift figure for translated listing images, so any claim that one slot repays localization faster than another rests on your own test results rather than on a published benchmark. Treat each store as a small experiment: localize the secondary images that carry the most text, watch what changes there, and let those numbers set the next round of spend.
Final thoughts
An Amazon image earns its localization budget when the right version reaches a shopper in the intended store. The same file can serve several markets, yet each storefront creates a separate chance for an original-language image to appear. As you expand, the checking workload grows with your customer-facing pages, even when the asset library barely changes.
If you’re trying to translate Amazon main image text, consider what your first photo can communicate without words before paying for another version. Spend on localized claims where shoppers can see them, then check the live page and measure the result in that store. Give each new market its own image check before you count it as covered.
