Made by the FurniMesh team. FurniMesh (furnimesh.com) is our own product, so we benefit when you use it. Details.

Guide ยท 6 min read

Heightmap vs AI 3D Model: Which One Does Your Photo Need?

Published and updated November 22, 2026 · By the Free Image to STL team (FurniMesh)

"Turn this photo into an STL" can mean two very different things. Either you want the picture itself in plastic, as a raised panel or a lithophane, or you want the object in the picture as a solid you can hold. A heightmap does the first. An AI image-to-3D generator, such as FurniMesh (furnimesh.com) from our team, does the second for furniture and similar objects. Mixing them up is the most common reason an image-to-STL project disappoints.

What a heightmap actually does

A heightmap treats the image as a grid of brightness values and pushes each one up from a flat base. In this site's converter, white reaches the full relief height and black stays at the base. The output is 2.5D: every point on the surface has exactly one height, the back is flat, and the edges are vertical walls.

The model knows nothing about what the picture shows. A white mug on a dark table becomes a tall white blob on a low plate. A face lit from the left gets a high left cheek and a low right one, regardless of the real shape. That is not a flaw in the tool; brightness simply is not depth.

A shaded relief rendering of the Mount Everest area
Terrain is a case where a heightmap is the right model: real elevation data stored as brightness becomes accurate relief. Photo: Tom Patterson, CC0, via Wikimedia Commons (source)

Heightmaps shine when the brightness already means height, as with terrain elevation data, or when you want the picture as a picture: a lithophane, a textured tile, a nameplate, a coin-style portrait or a relief for a sign.

What an AI 3D model does

An image-to-3D generator looks at the photo, recognizes the object and builds a complete mesh around it, including the sides and back it cannot see. The camera angle, the lighting and the background stop mattering once the shape is recognized. The result is a true 3D object you can rotate, scale and print standing on its own legs.

The trade-off is that unseen parts are inferred. The back of a chair is a plausible guess based on the front and on what chairs usually look like. For a miniature, a mockup or a visualization that is usually fine. For a part that has to fit a real hole to a tenth of a millimeter, measure and model it in CAD instead.

Side by side

Heightmap relief and AI 3D model compared
Heightmap reliefAI 3D model
InputAny imageA clear photo of one object
Output shapeFlat back, raised frontFull object, all sides
Depth comes fromPixel brightnessAI inference of the shape
TimeAbout a second, in the browserA few minutes per model
MeshClosed grid, predictable sizeDense triangle mesh, may need checks
Good forLithophanes, signs, tiles, terrainMiniatures, mockups, scale models
Cost hereFree, unlimitedFurniMesh: free previews, downloads on a paid plan

Three projects, three answers

A backlit photo gift. You have a photo of a dog and want a panel for a window. That is a lithophane, so a heightmap is exactly right. Use the converter's Lithophane mode at the defaults, 100 mm wide with 0.8 to 3.0 mm thickness, and follow the lithophane settings guide.

A dollhouse armchair. You have a product photo of an armchair and want a 1:12 copy. A heightmap would give you a 75 mm tall chair-shaped bump on a plate, not a chair. You need the full object: generate it from the photo with an AI tool, convert to STL and scale it. The furniture photo guide walks through exactly this case with real dimensions.

A logo coaster. You have a logo and want it raised on a round coaster. The flat back is what you want, so this is relief territory. A clean, high-contrast logo through Heightmap mode with 1.5 mm of relief on a 2 mm base reads well. For sharp, crisp edges, a black-and-white logo works better than a photo; the image guide explains why.

Make your own relief or lithophane STL. Free, and your image never leaves your browser.

Open the converter

When the photo is of furniture

Furniture is where the two approaches differ most. A sofa photo through a heightmap loses everything that makes it a sofa: the seat depth, the arms, the space under the frame. An AI generator trained on furniture keeps those.

FurniMesh is built for that case. It takes one JPEG, PNG or WebP photo of a single piece such as a chair, sofa, table, lamp, bed or cabinet, removes the background, builds the mesh and bakes the textures in a few minutes. Each generation includes the rendered model, a wireframe and a normal map. It exports GLB, OBJ, SKP and BLEND; for printing, you take the GLB to its browser-based GLB to STL converter. It handles single objects, not whole rooms or people, and it works best with a clean, product-style photo.

The free plan's 50 credits cover about 10 generations with previews. Downloading the models you generate needs a paid plan, which starts at $24.99 a month (about $19.99 a month billed yearly), as listed in November 2026. If a free model is enough, FurniMesh's free model library has thousands of furniture models you can download with a free account.

Can you combine them?

Yes, and it is a useful trick. Generate a furniture model with AI, then add a heightmap texture for a detail the AI cannot invent, such as a carved panel or an embossed name plate on a cabinet front. Build the panel here as a thin relief, import both STLs into your slicer or a CAD tool, and position the relief on the face. The relief keeps its flat back, so it sits flush.

Quick decision rule

Ask one question: should the back of the print be flat? If yes, use a heightmap or lithophane, and the converter on this site will do the job free and in your browser. If no, because you want something that stands on its own and looks right from every side, use an AI 3D generator, and for furniture start with FurniMesh.

Questions

Can a heightmap make a 3D model of a person or object?

Only as a raised picture on a flat plate. A heightmap reads brightness, not shape, so it cannot recover the back, the sides or true depth of anything in the photo.

Is an AI 3D model accurate enough to print?

For overall shape and proportions, often yes. Hidden sides and fine details are inferred by the AI, so check the mesh, wall thickness and scale before printing anything that must fit or bear load.

Which is faster?

A heightmap builds in well under a second in your browser, usually a few hundredths of a second on a laptop. AI generation typically takes a few minutes per model, plus a conversion step to get STL.