Why does the SVG from ChatGPT not work on my laser?
A language model writes an SVG the way it writes a sentence: as text, guessing coordinates. What comes out usually has no physical size, no separation between what should be cut and what should be engraved, shapes that do not close, and sometimes a photo embedded inside the file. It is a picture of a drawing, not a drawing a machine can follow.
The four failures, in order of how often they bite
- No physical size: the file opens at whatever DPI the importer assumes.
- One flat layer: nothing says which line the laser should cut and which it should score.
- Open paths: an outline that does not close does not free the piece.
- An embedded raster: the “SVG” is a photo in an SVG wrapper, and a cutter cannot follow a pixel.
What to do instead
Let a model draw the picture, which is what it is good at, and let code turn that picture into geometry. A tracer follows the edges of the image and produces real paths; then the layers, the millimetres and the checks are arithmetic, not a guess.
That is the whole pipeline MakerWizard runs: the image model draws, VTracer traces, and MakerWizard's own code assigns the jobs, sets the size and runs the check. A language model never writes path data in it.
If you already have one
Open it in the laser SVG validator: it will tell you which of the four failures you have, and whether the file can be repaired or is better re-traced from a picture.