Image to STL for 3D printing

“Image to STL” describes three different jobs. A logo that should stand proud of a plaque is a relief. A photograph that should appear when light shines through the print is a lithophane. A toy, a creature or a prop that needs a back, sides and a base is a full 3D model, and no amount of height-mapping will give a flat picture those surfaces. Choosing the right route first saves more time than any setting afterwards.

Pick the route that matches the object

Relief. Brightness becomes height above a solid base. Use it for lettering, logos, line art, coins, test patterns and decorative panels. The image-to-STL tool does this locally in your browser with width, relief height, base thickness, tone curve, smoothing and inversion controls.

Lithophane. Brightness becomes thickness, inverted, so dark areas block more light. Use it for photographs that will be backlit: a window hanger, a lamp shade, a night light. The JPG-to-STL tool starts from thin, inverted settings with a minimum-thickness control.

Full 3D model. The picture is a reference, not a heightmap. The Bake3D studio generates a textured mesh that infers the hidden sides from one image, lets you inspect it from seven views, and exports a geometry-only STL of the current pose from its export tab. That is a separate, credit-based process; the image-to-rigged-character guide covers image preparation and inspection.

If the object has a silhouette you would recognise from the side, it is a full model. If it lives on one flat face, it is a relief or a lithophane.

Prepare the picture

For a relief, prepare the image the way you want the height to read. High contrast between foreground and background gives clean walls; a transparent PNG background stays at the base thickness rather than becoming a raised slab. Anti-aliased edges become gentle slopes; increase smoothing slightly if JPEG noise turns into surface texture, but expect fine lettering to soften.

For a lithophane, tonal range matters more than resolution. A photo with detail in both shadows and highlights prints better than a flat one. Crop to the region you want, because the tool maps the whole visible image to a rectangle. Lighting in the photo becomes thickness, so a strongly lit face with deep shadows will print with those shadows as thick regions.

For a full model, use a single subject on a plain background, seen from the front or a three-quarter view, with the whole body visible. Occluded limbs have to be guessed by the generator, so a clean silhouette matters more than megapixels.

Set dimensions in millimetres

Both converters write STL in millimetres with a flat base at Z = 0, the convention slicers expect. Width sets the physical size; the image’s aspect ratio sets the other horizontal dimension. Resolution is sampled along the long edge, bounded to keep the browser responsive, so a 4,000-pixel photo does not become a four-million-triangle file.

For reliefs, start with a modest relief height. Small brightness differences become steep walls when the range is large. Base thickness is the minimum supporting material; pick it for your material and use rather than accepting the sample value as a universal profile.

For lithophanes, the minimum thickness sets the brightest possible region and the thickness range sets the contrast. Typical starting points are a minimum around 0.8 mm and a range of 2 to 3 mm, printed vertically with the picture facing out, but your filament’s opacity and your printer’s layer consistency decide the final values. Print a small test strip before a large piece.

For a studio model, the STL export carries the model’s scene size. Generated characters often use normalised dimensions, so read the reported size in your slicer and scale deliberately; a 1.8 m character becomes an 1,800 mm print at scale 1.

Check the STL before slicing

A successful download is not a printability certificate. Open the file in the STL checker and read the report. A relief or lithophane should report one shell, watertight, consistent winding and a positive volume; the converters build their meshes closed by construction and an automated test confirms that on generated geometry. A generated full model can legitimately report several shells if it has separate parts, but boundary edges, non-manifold edges or an inside-out verdict need attention in a modelling or repair tool.

Then check what the checker cannot see: wall thickness relative to your nozzle or resin process, overhangs, orientation and build-volume fit. Look at the sliced layers, not only the rendered preview. The rig and geometry checklist describes inspection habits for generated meshes that apply equally to print models.

Know the limits

A relief is a rectangle. Nothing in a heightmap can undercut, so a sphere in a photo becomes a dome and a figure becomes a raised silhouette. A lithophane reads lighting as thickness, so flash shadows and specular highlights print as geometry. Generated full models infer hidden surfaces and can get them wrong; thin parts, touching limbs and unusual anatomy are the usual trouble spots.

None of these tools repair meshes, check wall thickness, or know your printer. They convert and measure. The free converters run entirely in your browser with no upload and no account; images are limited to 32 MiB and 64 megapixels. The studio generator uploads your image to produce a model and charges credits; its costs and the refund-on-failure rule are on the pricing page.

References

The STL format description explains why the file carries no units and why slicers assume millimetres. The Blender manual’s STL import/export page documents scale and axis options if you want to edit the result before printing. For how the converters were tested, see the validation notes on the about page and the testing question at the end of each tool’s guide.