Text-to-CAD tools have often been used to produce one thing at a time: a bracket, a knob, a single enclosure. Version 2.0 of WeeklyVentures' browser-based modeler goes further. VoxelMatters reported on October 17 that CADlingo 2.0 is now in beta at cadlingo.com. It generates assemblies made of several separate parts, lets users inspect those parts one at a time, offers individual part export where export files are available, and outputs either STEP files for further CAD work or STL files for a slicer. According to VoxelMatters, the release comes about a year after the tool first launched.
For makers, this matters less as an AI headline and more as a workflow question: can a written description become a set of printable files without a trip through a traditional CAD package? This piece walks through what CADlingo says it does, where the multi-part change fits, and what to check before you trust its output on the print bed.
What CADlingo Actually Produces
The important word on CADlingo's product site is parametric. The tool takes a text prompt or a reference image and produces what WeeklyVentures describes as an editable 3D CAD model, not a fixed mesh. You can adjust dimensions after generation, refine the model through a chat interface, and restore earlier versions if a refinement goes wrong.
That separates it from the many AI tools that output a sculpted triangle mesh. A mesh looks fine in a render, but it is awkward to resize accurately: changing a bore from 8 mm to 8.2 mm to suit your printer's tolerances means editing geometry by hand. A parametric model keeps dimensions as values you can change. For functional prints, where clearances and fits decide whether a part works, that is the property that matters most.
According to VoxelMatters, generation, inspection, revision and export all take place in the browser, so there is no local install. The product site lists two export targets, STEP and STL, and pitches them for two different jobs: CAD editing and 3D printing.
The 2.0 Change: Assemblies, Not Single Bodies
The headline feature of version 2.0 is multi-part modeling. Rather than merging a request into one solid, CADlingo can now generate a model made of distinct components. VoxelMatters cites a bladed rotating disk, a bearing assembly with a shaft and housing, and a slider-crank mechanism as examples. The example prompts on CADlingo's own site include a bench vise built from six parts, a turbine disc and gears.
The companion feature is individual part export. Once an assembly is generated, users can inspect a single component and, according to VoxelMatters, download either the complete model or single components where export files were available. That matters for printing, because an assembly is rarely printed as one object. A slider-crank needs its crank, connecting rod and slider printed separately, each in its own best orientation, then assembled. A vise needs its jaws, screw and body on different plates, perhaps in different materials. Exporting one part at a time lines up with how makers actually lay out a build plate.
The choice of examples is telling. Vises, bearings, gears and crank linkages are all mechanisms where parts have to relate to each other: a jaw has to slide on its guide, a rod has to pivot on its pins. A tool that models those parts together, rather than as unrelated single bodies, appears to be aiming at functional assemblies rather than decorative objects.
A Description-to-Print Workflow
Based on the features WeeklyVentures describes, a typical session might run like this:
- Describe or show. Start with a text prompt or a reference image. Be specific about function and key dimensions. A prompt that names the mechanism and its overall size gives the generator more to work with than a vague noun.
- Review the assembly. Inspect each generated part individually. Check that the part count matches your intent and that nothing has been merged that should be separate, or split that should be one piece.
- Adjust dimensions. Use the parametric controls to set critical values: shaft diameters, wall thicknesses, overall envelope. This is where printer-specific clearance belongs.
- Refine through chat. For changes that are easier to describe than to type as numbers, use the chat refinement. If an edit makes things worse, roll back with version restore instead of trying to undo it by hand.
- Export by part. Export each component on its own where export is available, as STL for direct slicing or as STEP if you plan to finish the design in a conventional CAD program.
The STEP route deserves attention. STL is a triangulated surface: your slicer reads it happily, but most CAD packages treat it as a dead mesh. STEP carries solid geometry that tools such as Fusion, FreeCAD or Onshape can open and edit with real faces and edges. If the AI gets you most of the way and you want to add a fillet, a chamfer or a thread by hand, STEP is the format that makes that practical.
The Pricing Split
According to VoxelMatters, users can generate CAD models and export STL files for free, while STEP export requires eligible paid plans. The reports we reviewed do not specify plan names or prices, so check the pricing page on cadlingo.com for current terms before committing a project to the STEP route.
The split makes sense commercially. A hobbyist who just wants a printable part gets there without paying. Users who want to bring the model into a professional CAD workflow, which is arguably where a parametric generator adds the most value, are the ones asked to pay.
What It Means for Makers
Taken at face value, CADlingo 2.0 shortens the distance between an idea and a set of printable parts, especially for simple mechanisms that would otherwise take an evening of sketching and constraining in CAD. Multi-part generation and per-part export are the right features for that, because functional prints are usually assemblies.
Some caution is still warranted. The release is a beta, and nothing in the announcement covers dimensional accuracy, how the generator handles clearances between mating parts, or how reliably complex assemblies come out. Those questions decide whether a printed vise actually clamps or a printed bearing actually turns. Treat the output the way you would treat a model downloaded from a repository: measure the critical features, print a small test fit before committing to a long job, and expect to tune tolerances for your machine and material.
The sensible way to try it is on a low-stakes mechanism you understand well. Generate it, check the parts against your own expectations, adjust the dimensions that matter, and print the smallest mating pair first. If the parametric edits and per-part exports hold up, CADlingo could become a useful first-draft tool. Even then, it is best treated as a fast way to get a starting model rather than a replacement for knowing how your parts need to fit.
