Image / Text to 3D

TRELLIS-powered 3D asset generation

Tips for best results

For the best 3D model

One clear subject, centered and filling most of the frame.
Plain background. It's removed automatically, but a busy scene confuses the cut-out and bleeds into the shape.
Even lighting, sharp focus. Hard shadows and blur get baked into the geometry.
Higher resolution helps — a crisp ~1024px subject beats an upscaled thumbnail.
Avoid multiple objects — they fuse into a single model sharing one texture.

Drop an image here or browse PNG, JPEG or WebP, up to 20 MB

Tips for best results

For the best 3D model

Describe one physical object, concretely — e.g. "a weathered bronze dragon statue".
Keep it a short noun phrase, not a scene, story, or abstract idea.
Add material / style / colour cues ("low-poly", "matte ceramic", "rusted") for a better texture.
Text→3D is lower-fidelity than image→3D. If you have a reference picture, use Image mode instead.

0 / 1000

Examples

Recent

Source image

Model

Viewer

Scene Grid Background

Display Wireframe Normals Matcap Flat shading

Texture Channel

Base color

Metalness

Roughness

Specular

Normal

Generation

Sampling Seed Shape steps Shape guidance Detail steps Detail guidance

Input (image) Preprocess image

Mesh export Simplify Texture size Fill holes Hole size

Drag to rotate · scroll to zoom · right-drag to pan