3d-modeling

3D modeling

A production's film is a hint for video diffusion, which supplies the appearance, so each subject takes the least form that serves the hint. Read project context first for the representation the production adopted.

Default form

A subject the hint does not need takes no form and stays a description in the generation prompt. A person is the polygon man that entity work prepares. An object a person holds, uses, sits on or rides is an installed archetype where one offers it. Anything else a scene shows or touches is a coarse form, a few primitives that are just enough to tell what it is.

Requested detail

Go further only when the user asks for it, as a detail option, or when primitives cannot make a form identifiable. The user chooses its source: a diffusion model's output, code that shapes it in detail, or an existing mesh. When the user chooses a diffusion model, give it the subject's turnaround views, made under image as the model's request sets.

Build only what the request names and the subject's visible role needs. Finer geometry than that takes the time motion needed, and diffusion repaints it anyway.

Task	Read
Generate the form with a diffusion model	Diffusion
Shape it in code from dimensions and formulas	Crafting
Build or fit it from sections and guides	Lofting
Find an existing mesh, material, environment map or motion clip	Sources
Take a found or generated mesh into the production, or vary it by a parameter	Conditioning
Rebuild a fixed mesh as curves or surfaces	Recovery
Define parts, groups, placements or shared boundaries	Modularization
Add or change a skeleton, skin, attachment or deformation control	Rigging
Inspect a form's measures, coverage, construction or light	Inspecting a form

Whatever its source, a form is reproducible, rigged when it moves and judged on screen under visual debugging: a completed generation, fit, compilation or test establishes no visual result. The installed packages' documentation owns data formats and callable behavior.