Image to 3D Model: One Photo In, A Real Mesh Out
Upload a JPG, PNG or WEBP, generate a mesh from one view or from several angles, inspect it under ten shading modes, then export the format your next tool expects.
One image, or several
A single image is the fastest route and the one most people start with: the engine infers the back of the object from what it can see. It works best on a single subject on a plain background, lit evenly, with the object filling most of the frame.
When the shape matters — a face, a specific product, a part that will be printed — upload two or more photos of the same object from different angles. Multiview uses the extra views to reconstruct the sides a single photo leaves to guesswork, and it needs at least two images.
What makes a good reference image
One subject, plain background
a single object on a clean backdrop gives the segmentation step nothing to confuse it with.
Even, soft lighting
hard shadows get baked in as geometry. Diffuse light from two sides reads as shape, not as shadow.
Fill the frame
the object should occupy most of the image; a small subject surrounded by room loses resolution where it matters.
Avoid transparency and thin structures
glass, wire and cloth edges have no reliable silhouette, so reconstruction invents them.
Keep photographs out of the multi-view set
mixed lighting or a rotated background confuses the reconstruction more than it helps.
From upload to export
Pick a route, add images, choose a style preset and whether you want texture. Batch accepts a folder and creates one independent generation per image, each with its own progress and its own result — useful for a product range rather than a single hero asset.
Results open in the same viewport as everything else: ten shading modes from textured through wireframe, normal, matcap, ambient-occlusion-only, depth, UV checker and X-ray. The UV checker is usually the fastest way to spot a bad unwrap before export.
Picking the export format for your destination
| Destination | Format | Why |
|---|---|---|
| 3D printing | STL or 3MF | STL is read by every slicer; 3MF keeps units and colour |
| Web or game engine | GLB | one file with geometry, PBR materials and textures |
| Blender, Maya, ZBrush | OBJ | text-based and editable, materials via the .mtl side-car |
| Scan-style pipeline | PLY | carries per-vertex colour and custom properties |
Frequently Asked Questions
What image formats can I upload?
JPG, PNG and WEBP. PNG with a clean alpha channel helps when the subject sits on a flat background, because the silhouette is already isolated.
How many images does multiview need?
At least two, and more is better up to a point: three or four views spread around the object cover the geometry that one photo has to invent. Beyond that, returns drop and any inconsistent lighting starts to hurt.
Will a single photo give me a usable back side?
It will produce one, but it is inferred rather than observed. Symmetrical objects such as a mug or a chair handle it well; an asymmetrical face or a machine part does not. Use multiview when the back matters.
Can the result be printed?
Export STL or 3MF and check the watertight and manifold flags. A mesh generated from one view is often thin at the edges, so wall thickness may need a pass in the STL editor before slicing.
Do I have to run the whole batch again to fix one image?
No. Batch generations are independent tasks. Re-upload only the image you want to change and leave the rest of the batch as it is.
Are my uploads kept private?
New generations default to Private visibility, so the asset is visible only in your account. Public is an explicit choice, and you can switch it back at any time.