voice-to-3d-print

voice-to-3d-print, also called Pulsai 3D, is open-source, AI-first parametric CAD for turning an idea into an editable model and manufacturing files. The user starts from a description, a template or an existing CAD file, refines the result with parameters or conversation, inspects revisions and manufacturability signals, and exports STEP, STL, GLB, DXF or optional FDM G-code. The editable build123d and OpenCascade model is the source of truth, and meshes and previews are derived artifacts.

The README's status is public alpha. The deterministic CAD path works without paid AI, and a reproducible demo generated by a public GitHub Actions workflow creates the real parametric phone-stand model, applies a deterministic conversational edit, verifies that the revision and mesh changed, and prepares a print bundle. The hosted alpha at 3d.pulsai.app requires Google sign-in, and AI features normally require the user's own provider key unless account access was explicitly granted.

The tool covers five starter designs, text prompts, build123d source, STEP and STP import and STL fallback; parameter, named-feature, full-script, chat and optional voice edits; 3D preview, revision history, feature context and FDM and CNC manufacturability heuristics; and optional Anthropic, OpenAI, Gemini, Meshy, Tripo, Trellis2 and TripoSR integrations. The README states Pulsai 3D is not safety-certified and tells the user to independently review every model, slicer profile and G-code file before manufacturing.