An STL file does not contain units. It contains a list of triangle coordinates and nothing else — no "mm" marker, no scale, no defined size. Every slicer assumes the numbers are millimetres, and that assumption is the source of almost every "why did my model import at the wrong size" problem.
Why your model arrives at the wrong size
There are two opposite failures:
- It arrives huge (around 25× too big). The file was exported in inches but the numbers were written as if they were millimetres. 1 inch = 25.4 mm, so a 100 mm model imports at 2540 mm.
- It arrives tiny (10× or 1000× too small). The export used centimetres or metres — common with CAD tools and terrain data.
A quick diagnosis: measure the model on the build plate. If the number is roughly 25.4× what you expected, it is a unit mismatch, not a scaling mistake. If it is a round factor of 10 or 1000, the same — someone exported in cm or m.
Scale by target size, not by percentage
Every slicer will let you type a percentage, and percentages are how people end up with a 97 mm model when they wanted 100 mm.
- Cura — select the model, open the Scale tool, and switch from % to mm if you know the target. Lock uniform scaling before you touch a field.
- PrusaSlicer — right-click the object → Scale. Size fields are in mm, and a padlock icon keeps the proportions locked.
- Bambu Studio — the Scale tool shows X/Y/Z with a chain icon. Click the chain to lock uniform scaling before typing anything.
Pick your longest dimension, type the target, and lock the other two. That is the whole procedure.
Uniform versus non-uniform
Scaling all three axes by the same factor is safe for any model: it is just a bigger or smaller version of the same object.
Scaling one axis only is where prints break. A relief plate stretched in Z becomes a tall carving with impossible overhangs; a lithophane squeezed in Z loses the thickness range that encodes the image. Non-uniform scaling is for fitting a bracket into a slot, not for artwork.
Scaling a heightmap or lithophane
These models are special: one axis carries the picture, and a narrow band of thickness carries the tone. That creates two rules.
| What you change | What happens |
|---|---|
| X and Y, uniformly | Correct: same image, larger or smaller, same thickness range |
| All three axes by one factor | Thickness leaves 2.5–4 mm, so contrast collapses |
| Z only | Brightness mapping is destroyed |
| Print bigger without reconverting | Detail is stretched, not added |
So the correct habit is: if you want a 150 mm panel instead of a 100 mm one, do not scale the STL — reconvert the image at 150 mm. You get the detail your printer can actually resolve at that size, and the thickness stays where it belongs. Scaling up just makes every staircase step bigger.
The same applies in reverse. A 200 mm panel scaled down to 80 mm keeps the layer steps of the bigger print and loses the fine structure.
When scaling up is fine
Scaling is the right tool for mechanical parts and for ornamental models: keyrings, brackets, miniature furniture, a model that needs to fit a shelf. For those, uniform scale plus a quick check on wall thickness is enough — walls under about 1.2 mm get fragile when you shrink a model, and overhang angles get steeper when you enlarge it, so a model that printed without supports at 100% may need them at 200%.
Check the result before slicing
Drop the scaled file into the STL viewer. It reads binary and ASCII STL, reports the bounding box in millimetres and the triangle count, so you can confirm the size before committing to a long print. If the size is right and the preview looks solid, the mesh is fine.
If you are starting from a photo rather than an existing mesh, there is nothing to scale: set the width in the converter, the model comes out in millimetres, and the thickness stays in the usable range automatically.