One model, every product in the catalogue
AI-generated models are inconsistent by default: every generation is a new stranger, and a storefront where the model changes between listings reads as assembled rather than photographed. A model fixes one person — three angles, built once — and every later shot points back at the same person.
How it works
Watch the demo on YouTubeThree angles of one person
Front
Left 45°
Right 45°Product photos
FrontReference photos
Front
The same model, wearing different products
Product one
Product two
SceneWhy three angles and not one
One angle cannot hold a person. A single frontal photograph leaves everything about a face that only turning reveals unspecified, and each later generation fills those gaps differently.
Front plus a 45° turn each way triangulates it. That is what makes a shot generated next month recognisably the same person as one generated today.
Want a more accurate build? Add a full-body photo. Does your model often smile? Add a smiling photo. More photos help consistency; they do not guarantee it.
Two ways to build one
Upload three head-and-shoulders angles of a real person you are authorised to photograph (front and a 45° turn each way), with the short side at least 1024px and shot in consistent light. Full-body and smiling photos are optional extras.
Or describe the person you want in a short piece of text and have the three views generated. No photographs required. The optional full-body and smiling photos can be generated later, one step at a time.
Why consistency is worth the setup
Real photography carries consistency for free — one shoot, one model, one afternoon. Generated imagery does not, and buyers notice the difference before they can name it. A catalogue in which the face changes every listing reads as untrustworthy, and trust is upstream of conversion.
A model is a prerequisite rather than a product: build it once, then spend it across every submission that follows.
Frequently asked questions
What is a model?
Three head-and-shoulders photographs of one person (front and a 45° turn each way) stored together as a reusable anchor, optionally joined by full-body and smiling photos. Selecting it on a submission makes the resulting pictures show that person.
Why do AI models normally change between images?
Because each generation is independent. Without something anchoring the person, a model produces a plausible new person each time rather than the same person. Three fixed angles are what turns "a person" into "this person".
Can I use a real person as a model?
Yes, provided you are authorised to photograph and publish images of them. Upload three angles with the short side at least 1024px, shot in consistent light; full-body and smiling photos are optional.
Can I create a model who does not exist?
Yes. Write a short description of the person you want and the three views are generated from it.
Does the model I pick have to match the person in my reference photo?
No. When you pick a model, the person in the reference is replaced by yours (face, hair and build), while the pose, expression, framing, light and background stay as they were.