Choose how to create a 3D model
Start from what you already have. A photo or illustration belongs in Image to 3D. A written idea belongs in Text to 3D. Front, side and back images of the same subject belong in Multi-view to 3D. If you already have a GLB, STL or OBJ, open it directly; generating another mesh is unnecessary.
Make a 3D model from one clear subject
For an image, show the whole object against a simple background. For a prompt, specify the subject, proportions and important parts before adding material descriptions. A multi-object scene is harder to inspect and revise than one prop. Select the engine and read its availability and output options before submitting. A texture-capable model can generate appearance; a geometry-only model supplies shape.
Inspect the geometry before the surface
- Wait for the result to load in the main preview. Select a different result from the bottom thumbnail strip when several are available.
- Click X, Y and Z view points to check sides, top and front. Click a selected point again to see the opposite side.
- Switch to Clay to reveal the silhouette without textures. Switch to Wireframe to examine density and uneven triangles.
- Toggle the grid and axes as needed. Look for missing parts, disconnected pieces and surfaces that only look convincing from one angle.
Refine materials and export your 3D model
Open Materials & textures to adjust color, roughness and metalness. These controls edit the model locally; they do not run AI texturing. Applying an image map requires existing UVs or the simple box projection, which may show seams. Restore original materials if the edit loses useful detail.
Export GLB to retain supported materials. Use STL or this tool’s geometry-only OBJ when surface appearance is not needed. Reopen the file in its destination application and verify scale. Printability requires separate geometry checks. Read the 3D file formats guide.
Practice before using generation credits
Open the Gundam Mk-II example to practice viewing, material inspection and export with an existing licensed model. This is a viewer exercise, not an AI generation result. Generation time varies by service and queue; this guide makes no fixed “minutes” promise.