Convert a JPEG into an STL

What a photograph can honestly become

This is the part worth being clear about first, because it is what most people arrive hoping for. A photograph does not contain the shape of the thing in it. It contains how bright each point was. Those are different, and no tool anywhere can turn one into the other from a single picture — the information is simply not in the file.

So what you get is a relief: a surface whose height follows the brightness of the picture. On the right image that is a genuinely good result and looks like the subject. On the wrong one it looks like a melted version of the lighting, because that is what it is.

The images that work are the ones where light and shape already agree. A face lit softly from the front, a pet against a plain background, a silhouette, a signature, a mountain skyline. The images that do not are busy scenes, anything with a bright background behind a dark subject, and anything where a shadow falls across the thing you care about — the shadow becomes a valley cut through it.

Two things a JPEG brings that a PNG does not

The first is its own compression. JPEG saves space by throwing away detail in blocks of eight pixels by eight, and on a smooth area — a sky, a cheek, a wall — that leaves faint square edges. You have never noticed them on screen. In a heightmap they become real steps in real plastic, about a fifth of a millimetre across, and they catch the light.

The smoothing setting is the answer, and one is usually enough. It averages each sample with its neighbours, which removes the block edges and the camera speckle underneath them while leaving every edge you can actually see. Turn it up to two or three for a photograph taken in low light, which is where the speckle is worst.

The second is that a JPEG has no transparency. There is no alpha channel in the format, so there is nothing in the file that says where your subject ends and the background begins. If you want the shape cut out rather than raised, switch the first box to "the shape in it, cut out" and set the darkness threshold by eye — or save the picture as a PNG with a transparent background first, which is exact.

Why there is a base, and what happens without one

The darkest pixel in a photograph is black, and black means a height of zero. A surface built straight from the brightness therefore touches nothing at every dark point, and what you get is not a plaque with dark parts — it is a sheet with holes torn through it wherever the picture was dark.

The base thickness is the flat material underneath that stops it. One millimetre is enough to print and thin enough not to waste plastic; go to two or three if the piece is large enough to flex. The height of the relief is added on top of it, so a base of 1 and a relief of 4 gives a model 5 mm thick at its brightest point.

You can see this working in the measurements under the 3D view. The volume figure only exists for a closed shape, so a number appearing there is the proof that what you are about to download is a solid a slicer will accept rather than a surface it will refuse.

How big, and how much detail

The size box sets the longer side in millimetres and the other side follows the shape of your photo, so nothing is ever stretched. 100 mm is a plaque that fits on most print beds; 60 mm is a fridge magnet.

The detail box is separate, and it is the one that decides the file size. It sets how many samples are taken across the longer side, and every sample becomes geometry — 200 across is around 120 000 triangles, 400 is around 480 000, and the count beside the model tells you exactly where you are.

A big photograph is reduced to that grid before anything is built, and you are told when it happens. This is not a shortcut: a 4000 by 3000 photograph at one triangle per pixel is roughly 48 million triangles, which no browser will hold and no slicer will open. The reduction averages the pixels on the way down rather than dropping them, so detail becomes smoothness instead of becoming noise.

For printing, there is a floor worth knowing. A 0.4 mm nozzle cannot reproduce anything finer than about 0.2 mm, so on a 100 mm plaque a detail setting above about 500 is describing bumps your printer physically cannot make.

STL keeps the shape and nothing else

An STL file is triangles. It has no place to put a colour, a material or an image, so the photograph itself does not travel into the file — only the surface its brightness produced. This is the format, not a limit of this tool, and it is why the result is a relief you can feel rather than a picture you can see.

If you want the picture to stay a picture, STL is the wrong destination. Convert the same file to GLB instead and choose "a flat plane carrying the picture", which keeps every pixel as a texture. That is a different object for a different purpose: one goes to a printer, the other goes into a scene.