Turn a photo, sketch, or
technical drawing into a 3D model.
Upload a reference photo of a part to replicate, a hand-drawn sketch with measurements, or a dimensioned engineering drawing. Prompt2CAD reads the sizes and features it can see and builds editable parametric CAD — not a frozen polygon mesh.
120 free credits · No credit card required
Three ways to start
Anything you can capture as an image becomes the starting point for a real CAD model.
A reference photo
Photograph an existing part, fitting, or piece of furniture you want to replicate or replace. The agent reads the proportions, features, and visible dimensions and rebuilds it as editable CAD you can then adjust.
A hand-drawn sketch
Snap a photo of a napkin sketch or a notebook drawing with a few measurements jotted on it. The agent interprets the shape and the dimensions you wrote and turns them into a dimensioned 3D model.
A dimensioned technical drawing
Upload an engineering drawing with orthographic views and dimensions. The agent extracts the sizes, hole positions, and features it can read and builds the geometry to match — then you refine by conversation.
From image to editable CAD
What happens between your upload and your STEP file.
Upload your image
Paste or upload a reference photo, a hand-drawn sketch, or a dimensioned technical drawing straight into the chat. Add any context the picture can't show — material, hidden dimensions, how parts fit together.
The agent reads and plans
The agent analyses the image for proportions, features, and any written dimensions, then plans the structure before modeling. It asks for the dimensions and clearances it can't infer from the picture alone.
It builds real CAD geometry
Instead of tracing pixels into a frozen mesh, the agent calls geometry construction tools to build a parametric solid with explicit dimensions — the kind you can measure, edit, and export to STEP.
Refine and export
Inspect the model, click parts to check dimensions, and revise by conversation — "make the slot 1mm wider", "move the holes 5mm inward". Then export STEP, DXF, STL, or 3MF for your downstream workflow.
Why this beats image-to-3D
A mesh you can't edit versus geometry you can actually build from.
Editable geometry, not a frozen mesh
Image-to-3D tools trace your photo into a static polygon surface you can't dimension or edit. Prompt2CAD builds a parametric solid with real units, so you can revise a dimension and rebuild in seconds.
It reads the dimensions you give it
A photo alone gives proportions, not precision. Pair it with a few written measurements or a dimensioned drawing and the agent builds to those exact sizes — then you tighten tolerances by conversation.
It pairs with the tools you already own
Export STEP for Fusion 360, SolidWorks, or FreeCAD, DXF for AutoCAD, or STL and 3MF for your slicer. Use image-to-CAD to get to a structured starting point fast, then detail it in your usual editor.
No CAD experience required
Describe or upload what you have and the agent handles the geometry. You review dimensions in 3D and steer with plain-language changes — no sketches, constraints, or extrusions to define by hand.
What people build from an image
3D-printed replacement parts
Photograph a broken bracket, knob, or clip and get a dimensioned model to print. Refine the fit — "tighten the snap by 0.2mm" — then export 3MF or STL for your slicer.
Replicate existing furniture or fittings
Reproduce a shelf bracket, a leg, or a trim profile from a photo. Capture the proportions and any dimensions you can measure, then adjust the material and size to suit.
Mechanical parts from drawings
Upload a dimensioned engineering drawing and the agent builds the bracket, plate, or enclosure to the sizes and hole patterns shown — then export STEP for tolerance and fit work.
Frequently asked questions
Is image-to-CAD the same as image-to-3D?
No. Image-to-3D tools like Tripo or Hunyuan trace your photo into a polygon mesh: visually close, but dimensionless and hard to edit as CAD. Image-to-CAD builds a parametric solid with real units from the dimensions and features the agent can read, exportable to STEP and DXF for downstream CAD and CAM work.
Does it work from a photo with no dimensions?
Yes, but a photo alone gives proportions, not precision. The agent will estimate category-typical sizes and ask you for the critical dimensions it can't infer. For accurate results, add a few measurements in the chat or upload a dimensioned drawing alongside the photo.
What can I upload?
Reference photos of an existing object, hand-drawn sketches (a phone photo of a notebook page is fine), or dimensioned technical drawings with orthographic views. You can paste from the clipboard or attach image files directly in the chat.
How accurate are the dimensions from an image?
The agent is excellent at proportions and at reading dimensions you write on a sketch or drawing. Sub-millimetre tolerances and hidden geometry can't be read from a picture — provide those as text or refine them conversationally, and verify before manufacturing.
What CAD formats can I export?
STEP (.step) for solid CAD interchange, DXF (.dxf) for 3D geometry in compatible CAD software, OBJ (.obj) and STL (.stl) for mesh and 3D-printing workflows, 3MF (.3mf) for slicers, and GLB (.glb) for web 3D viewers. PNG viewport screenshots are also available.
Can I edit the model after it's generated?
Yes, in two ways. Conversationally — "make the bracket 10mm taller", "move the hole pattern inward" — and parametrically through dimension sliders the agent exposes. For deeper edits, export STEP and refine in Fusion 360, SolidWorks, or FreeCAD.
Keep reading
Start from an image
Upload a photo, sketch, or technical drawing and get an editable parametric model you can export to STEP, DXF, or STL. Free to try.
Upload your image →120 free credits · No credit card required