An STL is a surface: a shell of triangles with no colour, no material and no units. That is why "turn this picture into an STL" has more than one answer — what you are really asking is which surface you want, and there are three routes.
Route 1: convert the image to a heightmap (two minutes)
This is what most people actually need. The image's brightness becomes height: bright pixels stay low, dark pixels rise. The result is a solid plaque with the picture carved into the top.
- Open the converter and load a JPG, PNG or WebP file. It is decoded in your browser; nothing is uploaded.
- Pick relief mode, or switch to lithophane mode if you want a thin panel to backlight.
- Set the width in millimetres — the real size you want to hold. 80–120 mm covers most plaques.
- Set base thickness around 1 mm and relief height around 2–3 mm.
- Download the binary STL and slice it.
What you get is a closed solid: a flat base, vertical edges, and a height-mapped top. It is watertight and prints without repair in any slicer.
What you do not get is a replacement for a photo. The picture is now geometry — you cannot paint it back on, and the meaning of "dark" depends on the polarity you chose. The heightmap vs lithophane guide covers that choice.
Route 2: lithophane (backlit)
Same conversion, inverted polarity, plus a much tighter thickness window of 2.5–4 mm. Use it when the object will be held against a light. The lithophane maker sets the right defaults, and the night light guide covers the lighting side.
Route 3: trace it in CAD (real 3D geometry)
If the image is a logo, an icon or a letter, and you want an actual object — a keyring with a raised logo, a badge with a bevelled edge, a sign with mounting holes — you want CAD, not a heightmap.
The usual path: vectorise the image to SVG (Illustrator, Inkscape, or an online tracer), import it into Tinkercad or Fusion, extrude it to the depth you want, and add whatever holes or features you need. This takes an afternoon the first time and ten minutes after that.
| You want | Best route | Time |
|---|---|---|
| A photo or drawing as a plaque | Heightmap conversion | 2 minutes |
| A photo lit from behind | Lithophane conversion | 2 minutes |
| A logo as a raised keyring shape | SVG + CAD extrude | 20–60 minutes |
| A mechanical part with the shape as one feature | CAD | Hours |
The settings that decide whether it prints
Whatever route you take, these are what separate a print from a failed print:
- Minimum feature size. Anything thinner than about 0.8 mm will not survive a 0.4 mm nozzle. Fine hair, thin strokes and small text are the usual casualties.
- Vertical walls, not cliffs. A heightmap with 5 mm of relief in a 30 mm width creates near-vertical jumps between adjacent samples. More width or less relief fixes it.
- Resolution versus file size. A 4000 × 3000 photo becomes millions of triangles. Most printers cannot resolve that anyway — 200–300 dpi at the final print size is plenty, and keeps files under 20 MB.
- Base thickness. Under 0.8 mm the plate warps; 1 mm is the sweet spot.
- Units. STL has none. Slicers assume millimetres, so check the size on the build plate before slicing.
Common mistakes
- Converting a screenshot of a photo. It has already been resized and sharpened; converting it bakes in the compression. Use the original file.
- Expecting the converter to remove the background. It does not. A busy background becomes a busy relief; mask it out first.
- Printing a photo relief and comparing it to the photo. It is a carving, not a print — it will be read by shadows and it will look abstract from an angle.
- Trying to scale the STL afterwards to get more detail. Scale up a heightmap and you scale the staircase steps. Reconvert at the larger width instead; scaling an STL explains why.
Check the mesh before you print
Load the file into the STL viewer. It reads binary and ASCII STL, reports the triangle count and the bounding box in millimetres, and lets you spin the model. If the size is right, the shell is closed and nothing looks inside-out, you are ready to slice.
If you specifically want a 3D model rather than a relief, be clear about the difference: this workflow produces an image to 3D model conversion in the heightmap sense — a real solid, but one that only knows about the front surface you gave it.