KittyCAD, a software and API developer for hardware design, has announced the launch of its Zoo ecosystem, which includes a Text-to-CAD modeller and new machine learning API ML-ephant.
According to its creators, the goal for Zoo is to allow anyone to develop, use and customise modern hardware design tools, while also offering a selection of pre-built tools.
“When we had the idea for KittyCAD in 2021, we had only a preliminary idea of how our product roadmap would evolve,” said Zoo co-founder and executive chairman Jordan Noone, who also led Relativity Space’s technical development as CTO and co-founder for five years. “All we knew was that we wanted to build APIs to allow anyone to build modern hardware design tools.
“Fast forward to today, we’ve expanded our product roadmap to also include building certain tools on top of our own API.”
The Text-to-CAD modeller generates B-Rep surfaces, allowing 3D model to be imported and edited into any existing CAD software as a STEP file, compared to other text-to-3d examples that generate meshes that once imported can’t be edited in a useful manner.
“Generating CAD models is a lot different than generating images or video,” said Zoo co-Founder and CEO Jessie Frazelle. “Models generating 2D images, 2D video, and 3D point clouds are learning from datasets with dense, highly descriptive and strict representations of data that each have one and only one representation. In contrast, there are multiple valid feature trees for each CAD model, so training and validation are not straightforward.”
Frazelle added that in the future, Zoo will give users the generated KCL (KittyCAD Language) code so that it can be edited in its Modeling App. “Being able to edit code for a CAD model will make generated models even more useful,” she explains.
The newest addition to Zoo, ML-ephant, bridges machine learning and hardware design by shortcutting from design intent to a CAD model.
Starting with ML-ephant’s Text-to-CAD endpoint, designers can build prompt interfaces to generate starter CAD models through text prompts. The machine learning behind ML-ephant is trained on Zoo’s proprietary data sets and utilises KittyCAD Design API, which is built on Zoo’s custom, GPU-native Geometry Engine, to ‘programmatically analyse training data’ and generate CAD files.
“With the addition of ML-ephant and the creation of the Zoo ecosystem, we are building on the groundwork we’ve laid with the KittyCAD Design API to add machine learning capabilities to design tools never before possible,” continues Frazelle, who was previously CPO at Oxide Computer Company and a long-time infrastructure engineer.
Zoo’s proprietary Geometry Engine offers computation for geometry processing and visualisation rendering, allowing for tools to be built for any device or operating system.
DEVELOP3D readers can try out the Text-to-CAD tool for free here.
KittyCAD was previously part of the Nvidia Inception start-up program.
