Workflow Guide

How to go from a napkin sketch to a parametric model

ChatGPT (GPT-5) can generate the Python macro that becomes your fully parametric FreeCAD model. This page shows the repeatable loop that turns a notebook doodle into a manufacturable file.

Why this matters

No more line → constraint → dimension marathons. The macro is the model, so edits are made by changing text, not rebuilding geometry. You doodle like a human and author geometry like a compiler.

What you need

  • Sharpie or pencil + paper (front/top/side views if necessary)
  • A clear phone photo of the sketch
  • Description of function, tolerances, fabrication + assembly constraints
  • FreeCAD installed to execute the generated macro

The payoff

Generative → parametric → toolchain-safe → manufacturable, all in a single deterministic file. This is the workflow that will power open hardware, robotics, fixtures, jigs, and weird one-offs alike.

Step-by-step loop

  1. Sketch freely. Capture the idea with a bold marker so the shapes and dimensions are obvious. Annotate key constraints.
  2. Take a photo. Snap a straight-on picture with good lighting. Include multiple views (front / top / side) if the geometry warrants it.
  3. Describe intent. Tell ChatGPT what the part does, the mechanical tolerances you need, fabrication process, and any assembly constraints.
  4. Let GPT emit the macro. Paste the sketch photo + brief into ChatGPT. It writes the FreeCAD Python macro that defines the whole parametric model.
  5. Run it inside FreeCAD. Execute the macro. A fully constrained model appears instantly, ready for inspection, STL export, or tweaking.
  6. Iterate in text. Changing a dimension or adding a feature is editing the macro text—no more deleting half a sketch because a datum was missing.

Tips for reliable outputs

  • Be explicit about tolerances. Specify allowable clearance, press-fit targets, or "match M4 socket head" level details.
  • Call out fabrication + assembly. Mention whether you will 3D print, CNC, or laser cut, and any fixtures or hardware you plan to integrate.
  • Request parametric controls. Ask GPT to expose named variables for critical dimensions so you can tweak them later without rewriting logic.
  • Share results. Upload your macro, STL, or prompt to the community so others can remix the deterministic file.

This is the unlock

You move from doodling like a human to authoring geometry like a compiler. The feedback loop is short, deterministic, and production-ready. Once you try it, every lab tool, jig, or robotics bracket you make will follow this pattern.

Ready to try it?

Upload your macro, STL, or prompt to show the community what the new workflow unlocks.

Share an artifact