Skip to content

Use cases

AI for ecommerce

By Ofrai Editorial (Use Case Research)

Listings need stills, variants, and words. Ofrai keeps image models and writing models on one account so a catalog team is not juggling three separate logins.

Open studio

New accounts include 1 free GPT Image 2.5 generation. Other models and video use credits.

Creative demos

Give your product an opening shot

Explore an idea, then make it your own.

Sneakers through an hourglass

Water trails on a rainy platform

An underwater runway lights up

Creative demos. Results vary with your inputs and settings.

Product photography workflow

Start in Image Studio with Nano Banana Pro or Seedream for lifestyle lighting. If the card needs on-image type (sale badges, SKU names), A/B GPT Image 2. Keep backgrounds consistent in the prompt so a 20-SKU drop still looks like one brand.

Copy and PDPs

Draft titles and bullets in Text Studio (Claude or GPT), then generate alt-style hero shots without leaving Ofrai. Chat-with-PDF is useful when the source of truth is a spec sheet.

When to add a third model

Once you have locked the lifestyle lane on Nano Banana Pro and the typographic lane on GPT Image 2, add Seedream for a stylized variant. All three live in the same Ofrai image catalog and the same credits.

One balance across PDP and ads

Catalog drops use Image Studio for the lifestyle sweep and Text Studio for titles. PDP loops and ads reuse the same credits in Video Studio, so the budget is one number not three separate invoices.

Your first week: running a catalog sprint

Week one should end with a repeatable rule, not a folder of one-off images. Day 1: fix one hero setup (surface, light direction, lens) and write it as a prompt block you can paste again. Day 2: run Nano Banana Pro across your top 20 SKUs. Day 3: keep the three that landed, rewrite their prompts as templates where only the product name and colour swap.

Day 4: anything that needs letters on the card (a sale badge, a size chart) goes through GPT Image 2 with the same structure. Day 5: hand the catalogue owner the template folder, not the images. That folder is what makes week two cheaper, because the team stops describing light from scratch every morning.

Prompt patterns that survive a 20-SKU batch

A batch prompt is a template with three slots: product, material, background. Keep the slot order identical so two prompts can be read side by side and diffed. Example: 'Studio product photo of a matte ceramic bottle on travertine, 50mm, soft window light, shallow depth of field, no text.' You swap only the product and the material word, and the light stays on brand. Second pattern for flat lays: 'Top-down flat lay of {product} on {surface}, even soft light, no props, 1:1.'

Third for usage scenes: 'Lifestyle kitchen scene, {product} in use, morning window light, photoreal, unbranded packaging.' If a SKU needs something the other nineteen do not, give it its own prompt instead of bending the template.

Choosing an engine per shot type

Match the model to the job instead of fighting one. Nano Banana Pro handles lifestyle light and pack-on-set framing, which is most of a catalogue. GPT Image 2 handles the cases where letters must stay legible on the card: a sale price, a size chart, a variant label. Seedream 5 is your stylised and seasonal variant lane. Flux is for fast colour and layout passes when you are still deciding. The order of priority is realism first, legibility second, stylisation last.

A launch does not need a house style invented from zero: pick one engine per shot type, write the rule down, and let everyone else follow it.

Where product AI goes wrong — and the fix

Four failures show up in every launch. Drifting background: the model invents a new surface on each run, so name the surface and the light in every prompt and never paraphrase between SKUs. Wrong proportions: bottles get shorter and caps wander, so ask for a single object, no duplicate, and check the result against the spec sheet. Invented branding: a plausible-looking logo appears, so write 'unbranded, no label text' and route the output through legal before publishing.

Over-smoothed surfaces: leather and fabric turn to plastic, so add a material word such as visible weave or matte grain, or change engine rather than prompt.

Credits and margin math for a catalogue team

Treat credits as a cost per SKU, not as a monthly subscription. Before a drop, do the arithmetic: one hero per SKU, three variants for the top twenty, one card pass where text matters. Divide that job count by the balance you already hold and compare it against a single studio day. Text work is cheaper than image work, so titles and bullets should never be bought with an image model.

Writing the listing copy in Text Studio and generating the still in Image Studio is what stops the balance draining into retakes, and it makes the per-SKU cost predictable enough to quote a margin to a partner.

Who this is for — and who should skip it

It fits brands with 20 or more SKUs, frequent drops, and a team that does not want to manage three separate logins. It also fits small catalogues that need one presentable hero image before launch day. Skip it if your category demands a pixel-true packshot, such as jewellery or complex textiles: that needs a real photo set, and AI is only useful for the background and the campaign frames around it.

Skip it as well if nobody on the team reviews output before publishing, because Ofrai cannot absorb the legal risk of unverified model generations for you.

Deliverables the catalogue team can actually reuse

A generated image is not a deliverable until it survives the next channel. Before you save anything, fix three things. The file name: SKU, shot type, channel, version, never image2. The ratio ladder: export the 4:5 feed still, the 1:1 category tile and the 9:16 story version from the same prompt instead of cropping one into the others, because a crop loses the shadow the composition was built around. The written record of the prompt that produced the frame.

Keep that record next to your spec sheets in Chat with PDF so anyone can regenerate the shot when the campaign repeats next season. Then hand the catalogue owner two things: a folder of master stills, and one text file listing the prompts behind them, with each image labelled approved for ads, approved for organic, or mood reference only. That last label is what stops a stylised variant from shipping onto a price-sensitive marketplace listing by accident, and it takes ten seconds per image.

Getting started in one afternoon, in order

Begin in Image Studio with Nano Banana Pro, because it is the fastest lane to a result you can actually judge. Pick one product you know well, paste the studio product photo prompt, and generate four variants at the same ratio. Keep the one whose light matches your existing catalogue, then rewrite its prompt as a template.

Next open the same product in Text Studio and write a title, five bullets and a short description, pasting your spec sheet into Chat with PDF first so the facts come from your own document instead of the model's memory. If the listing needs a sale badge on the image, redo the hero in GPT Image 2 with identical wording and put the two side by side.

Credits are charged per job in the studio, so the cheapest way to learn the workflow is to batch your twenty strongest SKUs in one sitting rather than opening a fresh session per product. Everything you keep lands in your library, where it becomes the starting point for next season's drop.

Open studio

FAQ

Can I match a real packshot?
Use a tight prompt plus, when the model allows, a reference still. Always review legal and brand rules before publishing.
Is this a background remover?
Ofrai is a multi-model generator, not a dedicated cutout tool. Generate on a clean sweep in prompt instead.
How much does a generation cost?
Each job lists Ofrai credits in the studio. Catalog credits are starting points; resolution and duration can change the final spend.
How do I start on Ofrai?
Open Image or Video Studio from this page, pick a recommended model, paste or adapt the starter prompt, and generate from your library.
What is AI for ecommerce on Ofrai?
one place for product stills, listing copy, and short PDP loops: Nano Banana Pro and Seedream for lifestyle frames, GPT Image 2 for clean on-image type, Claude and GPT for titles and bullets.

Related models

Tools