ComfyUI lab 11: Design a text-to-3D handoff

Text-to-3D prompts should describe geometry, not a product photograph. “Beautiful cinematic radio” spends words on presentation. A useful asset brief names object class, major parts, proportions, symmetry, thickness, openings, support, and intended view.

Prompt for a mesh

Use one isolated object with a clear silhouette. Ask for watertight hard-surface construction when appropriate. Avoid transparent parts, hair-thin wires, dense text, and impossible intersections in the first pass. Generate broad form before surface ornament.

The open node sends the brief to the configured app.nz 3D endpoint, downloads the result, and saves a GLB into ComfyUI output. That is a handoff, not proof of production readiness.

Validate outside the beauty preview

Open the GLB in a neutral viewer. Check world scale, axis orientation, pivot, silhouette from six views, holes, disconnected components, self-intersections, triangle density, UVs, texture paths, and material compatibility. Rotate a hard light around it; shading artifacts reveal bad normals quickly.

Keep the prompt and source service/model with the asset. If the mesh will be redistributed, confirm both model terms and the rights to any references before publishing it.

Run the lab

• Open this graph in Comfy Studio
• Download the exact API-format workflow
• CLI: app comfy download appnz-text-to-3d

Record the graph JSON, input assets, seed where relevant, model or service version, custom-node commit, and final artifact together. That bundle—not a screenshot—is the reproducible creative unit.

Open-source map

• Open app.nz 3D node — MIT text-to-3D bridge
• TripoSR — Open single-image 3D reconstruction reference

Creative Systems Labs: 11 of 12. Previous: Restyle without losing structure Next: Turn a concept image into a 3D asset