Pic23D Open converter

Image to 3D model: pick the right route

Two very different techniques hide behind the same phrase. One turns brightness into height and keeps every detail; the other guesses the object and gives you depth. Here is how to tell which one your picture needs, and what neither of them can do.

Every "image to 3D model" tool on the internet is doing one of two things, and they are not interchangeable. Knowing which one you are looking at saves you an afternoon of bad results.

Route one: pixels become height

This is the oldest and most reliable trick. Each pixel's brightness becomes a height on a grid, the grid is closed at the bottom, and out comes a solid. Dark = thick, light = thin, or the other way round if you want a relief.

Nothing is guessed. Every detail in your original picture survives as geometry: a hairline in a logo, the freckles on a face, the grid of a map. What you cannot get is the back of the object — a photo of a mug becomes a curved slab shaped like the mug, not a mug.

Use it for: photos turned into lithophanes, logos and transparent PNGs, grayscale height maps, map tiles, stamps and cookie cutters. It runs inside your browser, so it is instant and unlimited.

Route two: a model guesses the object

Here an AI model looks at the picture, infers what the thing is, and generates closed geometry with a front, back and sides. A photo of a shoe comes back as a shoe you can spin around.

That guess is genuinely useful — and it is also the failure mode. The model cannot see what is not in the picture, so a side-on photo of a mug comes back as a mug with a plausible-but-invented back. Busy backgrounds, two overlapping objects and hard shadows are all read as real geometry.

Use it for: single objects you want as a mesh — a toy, a figurine, a statue, a product, a pet. It needs a server, so it has a daily free limit and takes a few minutes.

Choosing in ten seconds

Your picture Route
A photo of a person, for a backlit print Height map (lithophane)
A logo, PNG or line drawing Height map (plaque)
A grayscale map or a scan of a surface Height map (relief)
One physical object you want as a 3D shape AI generation
A scene with several objects Neither — crop to one object first

What neither route will give you

  • Texture. STL has no colour. If you need the photo on the model, you want a texture-carrying format (GLB) and software that can print or paint it.
  • Print-ready geometry. Both routes give you dense meshes. Scale in the slicer, and check that no wall is under about 0.8 mm before you commit to a two-hour print.
  • A perfect likeness. Height maps are exact but flat; AI is three-dimensional but interpretive. For a face you want to keep, the height map is the honest choice.

Sizes that work in practice

Lithophanes read best from about 80 mm wide upward, in white or natural filament at 0.12 mm layers, printed standing up. Plaques want a 0.6 mm minimum feature so small text does not weld itself together. AI meshes are usually scaled by eye: open them in the slicer, measure one known dimension, scale to it, then check the preview for detached parts.

Questions

Is a height map the same as a 3D model?

No. Height-map conversion keeps every pixel as geometry, but the result is flat-backed — you get a relief, not the whole object. Getting the back of a mug or a figure requires an AI model that infers the shape, which is a separate mode.

Can I print an AI-generated mesh straight away?

Usually no. AI meshes arrive dense, sometimes with thin limbs and occasionally as a shell rather than a solid. Open the file in your slicer, scale it, and read the wall-thickness warning before printing. Most slicers can repair small holes automatically.

Should I pick STL or GLB?

STL carries geometry only, which is what a slicer wants. GLB carries the texture as well, which matters if you are opening the model in Blender or a web viewer and want the photo on it. Choose STL to print, GLB to look.

What kind of photo gives the best 3D result?

One subject, filling most of the frame, evenly lit, against a plain background. That is the single biggest factor in how good any image-to-3D result looks — bigger than which model you use.