Newsletter
Stay Ahead
Occasional, engineering-led notes on applied AI — what we're building, running, and learning in production.
Occasional, engineering-led notes on applied AI — what we're building, running, and learning in production.
Use case: Generate product and lifestyle imagery without a photo shoot
Product photography is one of the slowest, most expensive parts of running an ecommerce catalogue: a proper shoot means booking a studio, a photographer, and often the physical product itself, then waiting days for edited files to come back. Webanto's image-generation use case replaces that bottleneck for a meaningful share of your catalogue — on-brand product shots and lifestyle scenes generated directly from a text prompt and, optionally, a reference image. Generation runs on flux.2-dev, a diffusion model Webanto self-hosts under a commercial licence on dedicated GPU hardware, called through POST /api/v1/images. Output is billed per image and scales with resolution, so a standard 1024x1024 hero costs a small, predictable number of credits — cheap enough to generate several variants per SKU and pick the best one instead of committing upfront to a single shoot result.
Faster turnaround from creative brief to usable image — no shoot to schedule, no studio dependency, no waiting on delivery of edited files.
Cheap enough per image to generate several prompt variants for the same product and pick the strongest one, rather than being locked into a single photographer's take.
Consistent visual style across a large catalogue, since the same prompt structure and reference image can be reused across similar SKUs.
Predictable, usage-based cost that scales with catalogue size rather than a fixed day-rate shoot fee, so smaller catalogues aren't paying studio minimums.
Generate a Webanto API key from Settings > API and store it as WEBANTO_API_KEY in your environment — never in client-side code.
Pick a pilot category (20-30 SKUs) and write a prompt template that fixes brand-consistent elements — lighting style, background, camera angle — leaving the product description as the variable slot.
Call POST https://api.webanto.com/api/v1/images with model set to flux.2-dev, your prompt, and 1024x1024 resolution to test the template against a handful of products before scaling up.
Budget the run: a 1024x1024 image costs roughly 5 credits, and larger formats scale at about 4 credits per megapixel, so 200 hero images a month at standard resolution runs 200 x 5 = 1,000 credits — comfortably inside the Growth plan's 50,000/mo allowance.
Route every generated image through a manual review step before publishing, checking for garbled text, anatomical errors, and anything that could be read as a real, specific product or person you don't have rights to depict.
Wire the approved images into your product pipeline — for example a webhook that fires on new-product creation — so generation becomes a step in the catalogue workflow rather than a manual one-off task.
Monitor credit usage on the dashboard and pick a plan tier that covers your steady-state monthly image volume plus headroom for seasonal campaigns.
You can steer flux.2-dev toward your product's shape, colour, and packaging using a detailed text prompt, but the model doesn't have your exact product in its training data, so results are an interpretation rather than a photorealistic copy. For anything that needs to match a real SKU precisely — exact packaging text, exact colourway — treat generated imagery as concept and lifestyle work, not a replacement for a photo of the actual item.
Generated people are synthetic, not real individuals, which avoids one set of consent problems, but you take on a different risk if the output resembles a real person closely enough to be mistaken for them, or if it depicts a real branded product other than your own. Review generated imagery with that in mind before it goes into paid ads or packaging, where the exposure is highest.
Image generation is billed per image and scales with resolution: roughly 5 credits for a 1024x1024 image, or about 4 credits per megapixel at other sizes. A credit is worth $0.01 at overage rates, so 1,000 images a month at standard resolution is roughly 5,000 credits — well inside the Scale plan's 200,000/mo allowance with room for other API usage.
Consistency across angles is the hardest part of prompt-based generation — the model doesn't retain a fixed 3D representation of your product between calls. Using a reference image as an input helps steady the product's proportions and colour, but expect some variation, and expect to be selective: generate several angles per product and keep the ones that match closely rather than assuming every generation will be identical.
For high-volume categories like commodity homeware, apparel basics, or seasonal variants, it can replace a large share of routine shoots. For hero shots, flagship products, and anything where photorealistic accuracy to the physical item matters — a phone's exact screen glow, a fabric's true texture — a real photograph is still the safer choice. Most catalogues end up using a mix of both.
Shoppers who spot a product in the wild, on someone else's feed, or on a competitor's site rarely know the right keywords to find it in your catalogue. Visual search closes that gap: a shopper uploads
Large media libraries accumulate redundancy fast — the same product shot re-exported at a different crop, a stock photo uploaded twice under different filenames, a banner re-saved at a lower compressi
Reverse image search answers a narrower question than catalogue browsing: given one photo someone just handed you — a screenshot, a competitor's product shot, a customer's snapshot of a damaged item —
Start a 14-day free trial on any paid plan. No credit card required.