Cartesian by Formas presents a three‑step workflow for generating 3D models: users describe the desired object using text or voice, provide a rough sketch or plan, and then drag in reference materials such as photos, scans or existing models. The AI interprets these inputs and produces a model where each component—like a chair, wall or table—remains a separate, editable part within the assembly.

Reference photograph of a skylit restaurant interior used as input for Cartesian by Formas AI modeling
The skylit restaurant reference photograph provided by users as one input to the AI systemwww.formas.ai

In the first step of Cartesian's workflow, users provide reference materials such as photos, scans or existing models for the AI to interpret and use as guidance when generating 3D models from text or voice descriptions and rough sketches.

Unlike image‑to‑mesh AI systems, Cartesian focuses on producing exact solids with clean topology, including native surfaces, precise edges and editable assemblies that avoid polygon‑mesh approximations. This approach allows users to inspect, measure and modify geometry directly, and it embeds semantic meaning (named elements and explicit relationships) to create a structured path into BIM, CNC fabrication or 3D printing workflows.

Exact 3D model of an interior kitchen showing separate editable components in red, pink, white and black colors
Interior kitchen exact model displayed as separate editable parts within the assembly after AI processingwww.formas.ai

Unlike image-to-mesh AI systems, Cartesian focuses on producing exact solids with clean topology where each component—like a chair, wall or table—remains a separate, editable part within the assembly, allowing direct inspection, measurement and modification of geometry.

Architecture Savoye red model demonstrating precise edges and native surfaces in Cartesian's editable solid output
Architecture Savoye red exact model showing clean topology with precise edges and editable assemblieswww.formas.ai

The tool produces exact solids with clean topology, including native surfaces, precise edges and editable assemblies that avoid polygon-mesh approximations, enabling users to embed semantic meaning for structured paths into BIM, CNC fabrication or 3D printing workflows.

The tool currently supports native export to Rhino’s .3DM and SketchUp’s .SKP formats, with announced future support for AutoCAD’s .DWG and the BIM‑standard .IFC format. Preview access began on September 18, 2026 via a waitlist; pricing and a general release date have not been disclosed. Demonstrated use cases span product design, furniture, interior architecture, urban design and manufacturing‑oriented outputs such as panel joints and CNC tool paths.