Product photography sits in a different category from marketing illustration or concept art. The fidelity to the actual product is non-negotiable: a generator that reimagines the product in flattering terms produces images you cannot ship. The discipline is closer to product visualisation than to creative generation, and the test protocols differ accordingly.
The product-photography priority stack.
- Photorealism. Lighting, material rendering, depth of field, shadow accuracy. Customers buy from photographic-looking imagery; illustrative renderings reduce conversion. Generators with strong material physics (translucent objects, reflective surfaces, fabric) and accurate shadow direction are the candidates. See /capabilities#photorealism.
- Product fidelity. The generated image must faithfully render the actual product. Geometry, material, colour, branding, packaging, all preserved. The mechanism is reference-image conditioning (uploading the actual product photo and prompting the generator to use it as the subject). Without rigorous reference handling, generators "reimagine" the product in ways that defeat the purpose.
- Background and lifestyle scene generation. Often the use case is replacing the studio backdrop with a context (kitchen counter, outdoor table, beach setting). The product is photographed once; AI generates the surrounding scene around the unchanged product. Tools that combine reference-image preservation with creative scene generation are the best fit.
- Batch generation at volume. A product catalogue with hundreds of SKUs requires automation. API access, batch endpoints, and reproducible parameters matter at this scale. Subscription platforms with manual workflows do not.
- Platform compliance. Where you sell or display the product image is governed by the platform's AI-content policy. Some require disclosure, some restrict by category, some are silent. The platform table on /licensing#platforms covers the main e-commerce platforms.
The standard workflow.
- Photograph the product once. Studio lighting, neutral background, front and three-quarter angles at minimum. This is the reference.
- Generate scenes around the unchanged product. Use reference-image conditioning to keep the product faithful. Generate variations of context (kitchen, dining room, outdoor cafe, lifestyle hands-on shot).
- Apply moderation and manual review. Check each generation for product fidelity, brand-safety issues, and accuracy. Reject and regenerate where needed.
- Apply human polish where it matters. A skilled retoucher in Photoshop addresses any residual fidelity drift, colour-corrects against the master, removes artefacts.
- Publish to your e-commerce platform. Confirm compliance with the platform's AI-content disclosure policy. Add disclosure in the listing copy where required.
What this replaces, and what it doesn't.
AI image generation handles lifestyle scene generation around a product economically. A photographer's studio shot remains the canonical asset; the AI workflow multiplies that one shot into a dozen contextual variations at a fraction of the cost of a multi-day photoshoot. What it does not yet replace reliably: hero macro shots requiring perfect material accuracy (jewellery, watches, optical products), products with intricate physical interactions (food in motion, liquid pours, fabric in movement), or any shot requiring real-world physics that the model cannot simulate accurately.
The economic case is strongest for e-commerce categories with broad product catalogues, frequent style refreshes, and lifestyle context as a key part of the listing, apparel, home goods, accessories, lifestyle electronics. The case is weaker for high-end goods where a photographer's eye for detail materially affects sale.
Platform policies (e-commerce-focused subset).
| Platform | AI-image stance |
|---|---|
| Etsy | Disclosure required for AI-assisted listings; some categories restricted. |
| Amazon (general retail) | AI-generated product imagery permitted with accurate representation of the actual product; misleading imagery prohibited under existing rules. |
| Amazon KDP (covers and interior) | AI-generated content disclosure required at submission. |
| Shopify | Platform-neutral; merchant's responsibility to comply with consumer-protection laws. |
| eBay | Misrepresentation rules apply; AI-generated lifestyle imagery is generally acceptable for showing context. |
| Walmart Marketplace | Listing content must accurately represent the item; AI-generated context permitted when product is faithfully depicted. |
Verify each platform's current policy on its own page before publishing at scale. The full table with links is on /licensing#platforms.
What to do this week.
- Pick three SKUs from your catalogue with strong base photography. Run them through two candidate generators (one closed-platform, one API-first) using reference-image features.
- Score the outputs on product fidelity, photorealism, and lifestyle context plausibility. Compare to your existing studio photography on conversion-rate-relevant attributes.
- Read the AI-content policy of every platform you list on. Note the disclosure requirements.
- If the pilot works, scale to twenty SKUs. Then build the API workflow for catalogue-wide automation.