AI Image Scaler: Keep the SKU, Upscale Only

An AI image scaler is a resolution pass, not a new pack shot. Upload the real file. Keep shape, color, label, logo, pack size, crop, and the field already in the photo. Change only pixel size: 2K for web and listings, 4K for campaign drafts, 8K only when you truly need a huge canvas. The result is not a lighting retouch, not a background swap, and not a restored photo of letters that were never readable. OpenAI’s Image API treats edits as a separate path from generating a new scene, and its size rules cap a single edge at 3840 pixels with a 4K pixel budget; resolutions above 2560×1440 are experimental. ChatGPT Images 2.5, launched 8 September 2026, is stronger at keeping a subject from a reference photo. Google’s Nano Banana 2.1 model page, dated 6 October 2026, lists 1K, 2K, and 4K output. FreeGPTImg is a GPT Image 2 workspace: 20 daily free credits, no credit card required. Overlay real words last.

Draft in FreeGPTImg. Run the keep-and-scale pass in the AI image scaler, polish light in the product photo enhancer only when that is a different job, cut a field in the AI background remover, keep materials honest in AI product photography, cut a marketplace white still in the Amazon product photo generator, change one noun from a locked hero in image to image, and stay on GPT Image 2 when that is the label on the screen.

What the job actually is

The file you already have is the SKU. Pixel count is the variable. If the bottle gains a new cap, a new flavor stripe, a marble slab, sharper invented type, or a second product, you did not scale the photo. You rolled a cousin.

OpenAI’s image-prompting guide is blunt about edits: start from the image you need, say what must stay the same, change one thing, then inspect. ChatGPT Images 2.5 is the consumer editing surface for that keep-subject loop. Google’s Nano Banana 2.1 docs describe multimodal generation and editing with stronger subject consistency and 1K / 2K / 4K output. None of that is a promise that a 600-pixel phone snap contains 8K of real detail. FreeGPTImg currently exposes GPT Image 2 on this loop. Write the name on the screen into the filename. Do not stamp a 2.5 or 2.1 badge on a GPT Image 2 file.

A scaler is also not a native-resolution generate. Generating a new 4K still from text is a different job: there is no locked upload. Scaling starts from a photo you already trust. Do not mix the two in one prompt.

Amazon’s product image guide still treats the main image as a realistic photo of the item you sell, on a pure white field, without extra graphics or type. A larger JPEG does not convert a lifestyle shelf into MAIN, and it does not waive a pixel-size rule in Seller Central. Read the live help article for the category you sell.

Name the scale before you generate

Weak files fail because they ask for 8K, a lighting overhaul, a new white sweep, a sale badge, and “restore the label” in one pass. Pick one job.

Job	What you export	Keep	Change one thing	Wrong move
2K web / listing	Same SKU, same field, larger file for store and social	Shape, color, label, logo, crop, background already in the file	Resolution to 2K	New lighting, new pack, extra props
4K marketing still	Same SKU, campaign-size pixels	The upload	Resolution to 4K	Fake micro-texture, rewritten type, a second SKU
8K print / poster draft	Same SKU, huge canvas for layout	The upload	Resolution to 8K	Treating invented grain as a true 8K capture
Lighting enhance	Different file, different tool	SKU and field	Light, sharpness, texture	Using the scaler as a retoucher
Background change	Different file, different tool	The SKU	Field only	Scaling and swapping the scene together
Native 4K generate	No upload, new still	Nothing from a SKU photo	Size at generate time	Skipping the real product file
Forensic restore	Do not export this	Nothing useful	Nothing honest	Asking the model to invent missing print

The scaler page on FreeGPTImg lets you pick 2K, 4K, or 8K. OpenAI’s documented GPT Image size envelope still tops out at a 4K pixel budget, with neither edge above 3840 pixels. If you pick 8K, inspect harder. A bigger file is not automatically a true 8K capture. Prefer 2K for storefronts and 4K for most campaign drafts.

Leave prices, ratings, and headlines empty if those words must be exact. Models still invent letters. Overlay in an editor after the SKU matches the upload.

Clean generation workflow

Use the same sequence in the AI image scaler, a GPT Image 2 tab, ChatGPT Images 2.5, or a Gemini session. The rule is the lock, not the logo on the chat.

1. Upload the real file. Front-of-pack, complete object, readable label, already at the crop you will publish. A tiny thumbnail, a screenshot of a competitor, or a photo of three SKUs is the wrong source.
2. Write five lines. SKU identity (shape, color, label words you must keep), the one scale (2K / 4K / 8K), crop, model name on the screen, and “no type.”
3. Write keep / change / do-not. Keep the upload, including the field and the crop. Change only resolution. Do not add a second product, a new backdrop, a logo you do not own, a person, or a price.
4. Optional look reference is look only. A second image may show how sharp a 4K still should feel. The product, logo, packaging, and field still come from the upload. Do not paste someone else’s listing as the hero.
5. Generate three options on the same crop and the same tier. Switching from 2K listing to 8K poster between option A and option B is a new job.
6. Inspect at 100%, then at the real placement size. Check melted labels, shifted logos, extra caps, plastic skin, fake pores, haloed edges, and unreadably invented type. If it fails at listing width, it fails in the store.
7. Route the file. Scaled listing, enhanced lighting, cutout, and Amazon MAIN are different downloads. Do not reuse one JPEG for all four.
8. Overlay real words last. Offer, date, and URL never belong in the generator if they must be exact.

If you are on Gemini, treat extra reference stills as the same SKU from useful angles, not as a moodboard of other brands. This job is one object and one resolution.

Prompt formula

Image scale only, not a new SKU, not a retouch, not a background swap.
Model: [GPT Image 2 / ChatGPT Images 2.5 / Nano Banana 2.1 / name on screen].
Source: uploaded photo of [one real product, pack size, label color].
Keep: exact silhouette, materials, label layout, logo, cap, colorway, crop, and the existing background.
Change only: output resolution [2K / 4K / 8K], same aspect ratio.
Text in image: none. Empty price, headline, badge, and watermark bands.
Do not add: extra products, people, pets, new backdrops, fake certifications, sale stickers, copyrighted characters, restored letters that were not readable.

Prompt examples

2K listing still from a locked upload

Keep this exact matte ceramic pour-over kettle, brushed steel handle, true product color, same spout and logo, same pure white field, same 1:1 crop.
Upscale only to 2K. Do not change lighting, materials, or pack size.
No new table, no extra kitchen tools, no steam, no type.
Do not invent a second kettle or rewrite the logo.

4K marketing still, same lifestyle field

Keep this exact kettle from the upload, including the wood counter already in the photo.
Change only resolution to 4K. Same crop, same light direction, same colorway.
No marble swap, no plants, no price tag, no “new” badge.
Empty labels. Same silhouette.

8K poster draft, inspect for invented grain

Keep this exact kettle. Scale to 8K for a print layout draft.
Do not add poster type, a headline band, or a barcode that was not readable.
Same field. Same crop. If detail was not in the upload, leave it soft. Do not invent pores or print.

Image-to-image, one noun: size

Image-to-image: keep this exact product, label, cap color, crop, and background.
Change only output size to 4K.
Do not add type, props, or a second SKU.

Review the pixels before you upload

Zoom to 100%. A thumbnail hides a melted barcode.

• SKU honesty. Same silhouette, colorway, pack size, and label. If a shopper would think it is a different flavor, discard it.
• Field honesty. The background in the upload is still the background. A new marble slab, beach, or loft is a different tool.
• Scale honesty. 2K, 4K, and 8K are output tiers, not proof of recovered optics. Soft print that was unreadable should stay unreadable or be overlaid later.
• Type honesty. Prices, stars, “new,” and URLs stay out of the pixels. Existing label words must match the pack you sell.
• Restore honesty. Blur that hid letters cannot become a new legal claim. If the print was unreadable, reshoot or overlay the real line in an editor.
• Rights honesty. Your product. No competitor packaging, no celebrity hands, no copyrighted holiday mascot.
• Route honesty. MAIN, gallery, ad, cutout, and scaled draft are different files. Amazon’s main-image rules still want a realistic product on pure white, not a larger lifestyle shelf.

Limitations you should not skip

FreeGPTImg is a GPT Image 2 workspace with 20 daily free credits and no credit card required. The AI image scaler is an edit. That is not a claim that FreeGPTImg runs ChatGPT Images 2.5 or gemini-nano-banana-2.1, that every small JPEG contains 8K of real detail, that missing print will be recovered, that you own commercial rights in every territory, or that a marketplace will accept the file. Match the label on the page you opened.

OpenAI’s current Image API documents recommended sizes at 1024×1024, 1536×1024, and 1024×1536, with custom WIDTHxHEIGHT strings whose edges are multiples of 16, aspect between 1:3 and 3:1, neither edge above 3840 pixels, and a total pixel count up to 8,294,400 (4K). Resolutions above 2560×1440 are experimental. Transparent PNG or WebP is a different export (background: "transparent"). Keep it off this loop.

Do not upload a scaled lifestyle still as Amazon MAIN because the kettle looks sharper at 200 pixels. Read Amazon’s product image guide for the category you sell. Leave type off MAIN. Keep sale treatments for other slots.

Models still invent extra objects when you ask for “more premium,” “gift-ready,” or “make it 8K luxury.” Shorten the brief. One SKU. One scale.

FAQ

Is an image scaler the same as a product photo enhancer?

No. A scaler keeps the field and the light, and changes only resolution. An enhancer keeps the field and changes lighting, sharpness, or texture. FreeGPTImg’s AI image scaler does not promise a retouch, a new backdrop, or a pixel-perfect transparent PNG.

Is 8K output real captured detail?

Not automatically. The scaler tool can request an 8K canvas. OpenAI’s documented GPT Image size envelope still sits at a 4K pixel budget, and Google’s Nano Banana 2.1 page lists 1K, 2K, and 4K. Inspect at 100%. If the upload was soft, a huge file with invented grain is a draft, not a capture.

Can I use the scaled file as an Amazon main image?

Only after you inspect it against Amazon’s product image guide. MAIN still wants a realistic photo of the item you sell, on a pure white field, without extra graphics or type. More pixels do not convert a shelf scene into MAIN. AI output is a draft until a person checks it.

Should I generate a native 4K still instead of scaling?

Generate native when you do not have a trusted photo. Scale when the SKU in the upload is already correct and you only need more pixels. Do not do both in one prompt.

Does FreeGPTImg run GPT Image 2.5 or Nano Banana 2.1?

Use the label on the page you opened. FreeGPTImg currently exposes GPT Image 2 on this loop. Write that name into the file. Do not stamp a 2.5 or 2.1 badge because a docs page used that id.

Can I upscale someone else’s listing photo?

Do not. Upload a photo you have the right to use. A competitor pack shot is not a sharpness reference. Look references, if you use them, control finish only.

Sources

• OpenAI: Image generation
• OpenAI: Image prompting
• OpenAI: Introducing ChatGPT Images 2.5
• Google AI for Developers: Nano Banana image generation
• Google AI for Developers: Gemini Nano Banana 2.1
• Google DeepMind: Nano Banana 2.1 model card
• Amazon Seller Central: Product image guide

Scale one SKU on FreeGPTImg

Write five lines: the real SKU, the one scale (2K / 4K / 8K), the crop, the model name on the screen, and empty type. Open FreeGPTImg, upload in the AI image scaler, polish light only in the product photo enhancer when that is a separate job, cut a field in the AI background remover, keep materials honest in AI product photography, cut a marketplace white still in the Amazon product photo generator, polish from a locked hero in image to image, and stay on GPT Image 2 when that is the loop in front of you. Download the image. Overlay the real offer last.