Image to STL: Full 3D vs Relief vs Lithophane

“Convert this image to STL” is incomplete until you decide what kind of third dimension you want.

A single image can become:

• a fully inferred object;
• a flat-backed silhouette extrusion;
• a grayscale relief;
• a backlit lithophane.

They use different geometry and solve different jobs.

Method	Geometry source	Best for	Biggest limitation
Full AI reconstruction	Semantic inference	Objects, characters, figurines	Hidden surfaces and scale are estimated
Silhouette extrusion	2D boundary	Logos, cutters, badges	Constant-depth shape only
Heightmap relief	Pixel brightness	Terrain, plaques, relief art	Flat-backed, brightness is not true depth
Lithophane	Thickness from tone	Backlit photographs	Must be printed and lit correctly

1. Full AI reconstruction

An AI image-to-3D model tries to infer the object represented in the picture. It estimates geometry beyond the visible pixels, may generate textures, and commonly produces GLB or another textured mesh before STL export.

Use it for a character, toy, shoe, chair, product concept, or pet figurine.

Do not confuse the output with a scan. The model cannot observe the rear, underside, internal structure, real-world scale, or manufacturing tolerance. Symmetric objects and clean three-quarter references tend to be easier than reflective, transparent, thin, or heavily occluded subjects.

Try Image to 3D Model when you want to retain a textured GLB, or Image to STL when fabrication is the destination.

2. Silhouette extrusion

Extrusion takes a closed 2D boundary and gives it constant depth. Imagine pulling a logo upward from a sheet.

This is the right method for wordmarks, signs, badges, cutters, stamps, tokens, and simple mounting shapes. It is predictable because the algorithm does not invent a back side or reinterpret the subject.

An SVG is normally the best input because its paths already define the boundary. Use SVG to STL for a clean vector. If you only have a raster file, PNG to STL can derive a shape, but anti-aliasing, compression, and a noisy background may create rough edges or unwanted islands.

3. Heightmap relief

A heightmap maps image brightness to depth. In a common configuration, black becomes the lowest point, white becomes the highest, and gray falls between them. The result usually sits on a flat base.

This is useful for terrain, coins, engraved plaques, embossed art, and depth-map experiments. It is not automatically a truthful reconstruction: a bright white patch in a photo may be a reflection, not a physically raised surface.

ImageToSTL's public explanation documents this brightness-to-height process and exposes inversion, base, smoothing, and detail controls.

Prepare a heightmap deliberately. Remove irrelevant background, use grayscale, control contrast, blur pixel noise, and preview both normal and inverted mappings before export.

4. Lithophane

A lithophane is intended to be viewed with light passing through it. Dark regions are usually represented by thicker material and light regions by thinner material, creating an image from transmitted light.

That makes a lithophane related to a heightmap but not identical to a generic relief. Minimum thickness, maximum thickness, curvature, orientation, material color, layer height, and the light source all affect the result.

Use a dedicated lithophane workflow when the goal is a backlit photograph. A generic image-to-STL tool may create visible relief geometry without producing the tonal behavior you expect under light.

A quick decision tree

Ask these questions in order:

1. Should the result have meaningful geometry around all sides? Use AI reconstruction.
2. Is the source a logo, word, icon, or outline with constant depth? Use SVG or PNG extrusion.
3. Should brightness control visible raised depth? Use a heightmap relief.
4. Should the image appear when light passes through the print? Use a lithophane workflow.

Why STL is usually the last step

STL is widely supported by slicers, but it is mainly a triangle surface format. A textured AI asset is often better preserved first as GLB, OBJ with materials, or another format that can carry appearance data.

Keep the textured source. Export STL when you are ready for a geometry-only fabrication workflow. This also gives you a better place to repair, scale, decimate, or edit the original asset before its print-specific conversion.

Inspect before slicing

For any method, verify:

• intended units and dimensions;
• manifold geometry and closed surfaces;
• duplicate, degenerate, or self-intersecting triangles;
• minimum wall and feature thickness;
• face normals;
• a stable base or bed contact;
• overhangs and supports;
• triangle count and slicer performance.

SVG Genie's automated 3D checks currently verify the GLB handoff, storage, and access-refund path. They do not yet certify these print-specific properties. The 3D Quality Standard publishes the exact boundary.

Final rule

Use semantic AI only when you need semantic inference.

For an object or character, reconstruct it. For a logo, extrude the vector. For depth art, use a heightmap. For a backlit photograph, make a lithophane.

The best converter is the one whose geometry matches the physical result you actually want.

Open all SVG Genie 3D workflows