An online store doesn't need one product photo. It needs many. The same shoe on white for the category page, on a table for the newsletter, in hand for the ad, in a setting for Instagram. Traditionally that means a photographer, a studio, styling, and retouching — tens of euros per image and weeks of turnaround. For a store with fifty products, it becomes a real obstacle: most shops end up using a single flat product shot everywhere.
Phora inverts that. You upload an existing image and get a ready-to-sell scene back in ten to thirty seconds — studio, lifestyle, flat-lay, macro — no shoot required. It's my own product; here I'm writing about the actual hard problem behind a tool like this, and the product decisions it forced.
What it is, and who it's for
The entry point is deliberately mundane: a JPG, PNG, or WebP up to 10 MB — the photo that's already sitting in the store or on a phone. The output, though, speaks the store's language: a 2K image (2048px), optimized as WebP, so the finished scene doesn't need another compression pass before it goes live.
And you don't have to think image by image. Phora processes an entire catalog in one run — hundreds of products in minutes — because a real store doesn't have one nice photo, it has a database of products. The interface runs in English and Hungarian; the audience is the store for which the photographer–studio–retouching chain never fit the margin.
The hard part isn't the generation
Calling an image generator is easy. The hard part is making sure the product doesn't change. A beautiful scene is worthless if the logo blurs, the label text gets rewritten, or the toe of the shoe takes on a different shape. The customer receives what they saw in the photo — and if the photo lies, that's returns and complaints. In product photography, "almost identical" isn't an acceptable result; it's a defect.
So Phora doesn't "paint a new image." It places the existing product into a new environment: the Google Gemini model preserves the labels, textures, and geometry from the uploaded photo and builds a scene around it. The difference is invisible to the user, but it's the whole reason the output is usable at all. Generation is the flashy part; product identity is the product.
Two routes: trust the machine, or take control
Not everyone wants to work the same way, so there are two modes. In smart mode the system looks at the product, suggests a fitting scene, and explains why — for someone who doesn't know photography, that's enough, and they pick something up along the way. In advanced mode you write the prompt, set the aspect ratio, and can supply a reference image when the output has to match an existing visual style — for someone who knows exactly what they want, the control is there. Same engine, two thresholds.
The two aren't tiers, they're ways of working: smart mode isn't a dumbed-down version, and advanced mode isn't an experts-only wall. You can prompt your way through a campaign and leave the rest of the catalog to the machine.
Nothing goes live without approval
AI gets it wrong sometimes, and the product doesn't hide that — it designs around it. Every generated image passes through an approval step: the old and new images side by side, and you decide what stays. Anything you don't like regenerates with one action; what passes moves on; what doesn't never appears in the store.
The whole flow runs from the keyboard — accept, regenerate, reject, step through — because for anyone processing five hundred products, speed is a feature: approval can't be slower than generation. The same principle runs through my other product, titkar.ai, just on invoices instead of images: the AI proposes and prepares, the final word is yours. When a tool publishes in your name — into your store, toward your customers — approval isn't a convenience; it's the foundation of trust.
It connects where the store lives
Approved images land on the right product in the store by themselves — no downloading, renaming, re-uploading. Phora talks to Shopify, WooCommerce, BigCommerce, and the Hungarian Shoprenter, and the connection follows whatever each platform allows: app install, OAuth, or API credentials.
It works without a store too: drop in an image, transform it, download it. The store integration is an accelerator, not a requirement.
Credits that don't expire
Pricing is credit-based: one credit, one image. The entry point is free, with ten images — enough to see what the tool does with your own products, not a demo's. Above that, credit packs carry volume discounts, and for anyone working at scale there's a monthly Studio plan with unlimited connected stores.
Two decisions matter more than the numbers: unused credits never expire, and they're refundable. A tool you reach for occasionally — a new collection, a season change — shouldn't punish you for resting between campaigns.
Design as trust
A tool someone uses to present their own products can't shout. The interface is dark, calm, and the color comes from the product photos themselves — the emphasis is on the image, not the frame. The landing page doesn't promise, it shows: real before-and-after pairs made with Phora, all the way down, demonstrating what it does. For a product like this, the best argument is the result itself.
Where it stands
Phora is live and usable: the free entry covers the full journey from upload through approval to an image synced into the store. It keeps growing — the integration list and the scene vocabulary along with it. The lesson is the same one my other AI products taught me: calling the model is an afternoon; the product is everything around it — the unchanged label, the approval step, the store sync, and pricing you can trust.
There's always a next level.
If you like what you see (whether you're building a product or a team) I'd love to hear about it.