Docs
Intelligence
The platform turns one photo into a full 3D model through several AI stages that run automatically. You don't pick engines or modes; there's one streamlined pipeline. This page explains what each stage does.
The pipeline
Every generation runs through the same stages, automatically:
- 1. Image enhancement. Your photo becomes a studio-lit, isolated, well-framed picture. Backgrounds are removed, lighting is normalized, the subject is centred. How much of the subject belongs in frame comes from the kind you picked: a whole figure is completed head to toe, a body part is left as just that part.
- 2. Pose + descriptor analysis. A vision pass extracts the subject's pose, outfit, style, and orientation. This metadata fills in the model's listing fields automatically (title, description, tags).
- 3. 3D generation. That cleaned-up picture (your own photo, not a redrawing of it) is fed to the 3D generator. The mesh comes back in glTF binary (GLB) and is stored privately in your account.
Why one picture, not four
It is tempting to assume more views means a better model. In testing, the opposite held: extra views that our AI invented disagreed with each other (a sleeve in one, a bare arm in another), and the generator averaged the disagreement into a worse mesh than one good picture produced alone.
So an automatic job sends exactly one image, the cleaned-up version of your own photo, and we spend the effort on making that one right instead. Several real photographs are different: those agree, because they are of the same real thing, and the step-by-step path is where you supply your own.
Colour is always included
There's nothing to choose here. Every model is built with full PBR textures (colour, normal, metalness and roughness) baked in at no extra charge, because our generator includes them in the base price.
If you'll prime and paint the print by hand, the colour simply doesn't get in your way: the geometry is identical either way, and your slicer ignores the textures. If you want the model to look finished as-is, for a render, AR, a display piece, or a colour-capable printer, it's already there.
What about your data?
We don't train AI models on your uploads or generations on any plan. The list of which AI providers handle which stage, and what each receives, is in Privacy Policy section 4.