Image to 3D: one picture in, a full model out

The generator can only be as accurate as what it sees, so the photo matters more than any setting. The short version:

• One object, fully in frame, with a little space around it.
• Even, soft light. Hard shadows get baked into the texture as dark paint.
• A three-quarter angle shows more depth than a flat front shot.
• At least 1024 px on the short side. Blurry inputs give soft geometry.
• Avoid strong perspective from wide-angle phone lenses close up; step back and zoom in.

There is a full walkthrough with examples in choosing a reference image for image to 3D.

A single image forces the generator to invent the back of the object. For symmetric products like bottles, lamps and vases that guess is usually right. For anything with a different back, such as a backpack, a chair or a character with a tail, upload extra views.

Views should show the same object under similar lighting. Label each upload as front, left, right or back in the uploader so the views are not mixed up. Multi-view generation costs the same 30 credits as a single image.

Clean line drawings and flat-shaded concept art work well, especially for stylised props and characters. Two things help: draw the object in a neutral pose on a plain background, and add a short text hint (for example “wooden treasure chest, iron bands”) so the generator knows which materials to paint.

Rough thumbnails with overlapping strokes tend to produce lumpy geometry. Clean up the outline first, or use text to 3D and attach the sketch as a style reference instead.

Only upload photos you took, artwork you made, or images you are licensed to use. Turning someone else’s product photo or character art into a 3D model does not remove their rights in the design. Uploaded images are private to your account and are not used for training unless you opt in. See model licensing and the acceptable use policy.