Convert a PNG into an STL

Transparency is what makes a PNG better than a photo here

A PNG can record, for every pixel, how see-through it is. A logo saved properly is therefore marks on nothing at all, and the file already states exactly where the shape ends — no guessing, no threshold, no halo of background left around the edge.

The cut-out mode uses that directly. Anything see-through is left out, the boundary of what remains is traced as a closed outline, and the outline is given a thickness. A letter O comes back with the hole in the middle, because the tracing works out which rings sit inside which. Nothing about that is special-cased for letters; it falls out of the nesting, exactly as it does for a washer or an island in a lake.

If your PNG has a white background rather than a transparent one, the same mode still works — it falls back to darkness, and the threshold setting is where you put the line. A black-on-white drawing separates cleanly. A photograph does not, and the raised surface is the right choice for one of those.

Lossless, which matters more than it sounds

PNG stores exactly the pixels it was given. Nothing is approximated and nothing is thrown away, so there are none of the faint block edges that JPEG compression leaves behind on smooth areas — the ones that turn into real steps in a printed surface.

That is why a screenshot, a diagram, a QR code, a line drawing or a piece of clip art should always reach this page as a PNG. Those images are large flat areas with hard edges, which is the exact case JPEG handles worst and PNG handles perfectly. You can usually leave the smoothing at zero and keep every edge sharp.

For a photograph, none of this applies and the file is simply larger than it needs to be. Photographs belong in JPEG, and there is a page for those.

Choosing between a raised surface and a cut-out

Ask what the dark part of your picture is. If it is a *shape* — a logo, a letter, a silhouette, an outline — you want the cut-out. You get a flat solid with that shape and a clean vertical edge, which prints quickly, needs no supports, and can be glued, inlaid or used as a stamp.

If the dark part is *shading* — a face, a landscape, anything with gradual tones — you want the raised surface. Cutting out a photograph gives you a meaningless blob, because a threshold has to draw a hard line through soft shading and there is no right place to draw it.

Both are in the first box under the 3D view, and switching between them rebuilds the model in front of you. That is usually faster than reasoning about it.

Thickness, and what makes it printable

A cut-out has one thickness and that is all: 3 mm is a sturdy keyring or fridge magnet, 1 mm is a stencil or an inlay, 10 mm is a shape that stands up on a shelf.

A raised surface has two numbers instead. The base is solid material underneath the whole thing, and it is not optional — the darkest pixel is height zero, so a surface with no base is a sheet with holes torn through it wherever the image was dark. The relief is how far the brightest part stands above that base.

Either way, the volume figure under the model is the check. Volume can only be worked out for a closed shape, so a number appearing there means what you are about to download will slice as a solid rather than being refused as a surface.

What reaches the STL file

The shape, at the size you set, and nothing else. STL holds triangles — no colour, no material, no image — so a red logo and a black one produce the same file. If the colours matter, convert to 3MF instead, which records its own units as well.

The image itself does not travel either. If you wanted the picture to stay a picture on a flat plane, STL cannot do it and GLB can: convert the same PNG to GLB and choose "a flat plane carrying the picture", which keeps every pixel, transparency included, as a texture.