Disclosure: I run VELIN, a small pay-per-image API, and I link to a prompt page on its site at the end. The prompts below work with any Nano Banana / Gemini image model.
If you sell on Etsy, Shopify or Amazon, you probably don't need a "better model". You need a repeatable way to get the same product into ten different scenes without the label, shape or colour drifting. This is the structure I use.
The one rule: separate "keep" from "change"
Every prompt has two parts:
- Scene sentence – the only thing that changes between runs.
- Keep sentence – the same text every time, plus 2–3 real reference photos of your product from different angles.
A keep sentence that works well:
Keep the product exactly as in the reference photos: same shape, proportions, label text, colours and materials. Do not add logos or text that are not on the product.
Change only the scene sentence between runs. If you change the scene, the lighting and the camera in one go, you can't tell what broke the product.
12 scene sentences
Replace {product} with a short phrase such as "ceramic coffee mug, matte sage green, no logo".
- White background main image – "Professional e-commerce photo of {product}, centred on a pure white background, soft even studio lighting, faint natural contact shadow, product fills ~85% of the frame, no props, no text."
- Lifestyle table scene – "{product} on a light oak table near a window, soft morning light from the left, shallow depth of field, two simple props that don't hide the product, 50mm look."
- Flat lay – "Top-down flat lay of {product} on neutral textured paper with 3–4 small complementary items, diffused light, generous negative space."
- Hands in use (no face) – "Close-up of a person's hands using {product}, no face visible, natural home setting, soft window light, realistic skin."
- Worn by a model (apparel) – "Full-length catalogue photo of an adult model wearing {product}, plain light-grey backdrop, soft box lighting, neutral pose, natural fabric folds."
- Gift-box / unboxing – "{product} with a simple unbranded kraft gift box and tissue paper, slightly angled view, soft side light, tidy unboxing look."
- Seasonal – "{product} as a gift in a cosy seasonal setting, warm fairy lights blurred in the background, plain ribbon, product clearly in the foreground."
- Macro detail – "Macro close-up of {product} showing material texture, very shallow depth of field, crisp focus on the most interesting detail."
- Outdoor – "{product} on a mossy stone outdoors, golden-hour backlight, soft green bokeh, realistic scale and shadows."
- Dark premium – "{product} on a dark reflective surface, dramatic rim light, subtle reflections, charcoal background, luxury ad look."
- Wall-art mockup (digital-download sellers) – "A framed print on the wall of a bright modern living room, soft daylight. The print shows: {artwork description}."
- Social ad with copy space – "Vertical ad image of {product} in the lower third, large clean empty area above for a headline, simple gradient background. Do not render any text."
Cheap drafts, expensive finals
Iterate at 1K, and only re-run the approved scenes at 2K/4K. On per-image pricing that is the same price at any resolution (on VELIN it is about $0.037 per image whether 1K, 2K or 4K), so draft and final cost the same and you can simply pick the size you need. On token-billed APIs, resolution changes the price, so drafting small matters more.
A quick budget for one listing: 10 final images × 2 tries each = 20 generations ≈ $0.75 at $0.037.
Things sellers get wrong
- Don't replace the real product photo. Use generated scenes around the real product and keep an honest photo of the actual item in the listing. Buyers (and marketplaces) care that the item matches.
- Disclose AI where the platform asks. Etsy has an AI-generated attribute in listing details; check each marketplace's current rules.
- Check labels and small text on every output. If text drifts, add the exact text in the keep sentence or composite it afterwards.
- Download what you want. Many services (VELIN included) auto-delete outputs after a few hours.
Try the presets
I put the 12 prompts into a small page where you type your product once and copy each prompt or open it in the studio: https://72agi.com/product-photos.html . There's also a cost calculator ( https://72agi.com/calculator.html ) and a copy-paste API script generator ( https://72agi.com/api-quickstart.html ) if you want to batch it from a spreadsheet.
VELIN is an async REST API (about 45–100 s per image; not OpenAI-compatible). Top-up is USDT only for now. If you'd like to start with a small top-up, message us on Telegram @ayan8866e.
Top comments (0)