Shape, topology, UVs, rigs and delivery formats

Suggested reading time: about 8 minutes. This is an editorial production guide. Generation, performance and pricing comparison experiments have not been run.

Learning goals

  • Distinguish a model’s shape from its rendering.
  • Understand the uncertainty in reconstructing 3D from an image.
  • Know what to check when delivering an asset for the Web.

Inspect the asset across the export boundary

  1. 1Required views and motion
  2. 2Blockout
  3. 3Topology, UVs and needed rig
  4. 4Delivery export
  5. 5Reload in the target runtime
  6. 6Inspect shape and materials
Consider the sequence and each role.

This original handoff diagram places a separate acceptance check after export. The editable authoring file and the runtime asset serve different purposes; unsupported features must be resolved before delivery.

Who does this work?

  • Modeler: shape and the structure of faces.
  • Rigger: deformation and controls.
  • Technical artist: export and runtime performance.

Inputs

  • Use case, required viewing angles and size.
  • Reference drawings and a specification for parts that are not visible.
  • The runtime’s supported format and target file size.

Production workflow

  1. Block out the shape with spheres and boxes.
  2. Check proportions from the front, side and back.
  3. Create the necessary details, UVs and materials.
  4. Add a rig or variants only to things that need to move.
  5. Export to the delivery format and inspect it again in the consuming application.

Outputs

  • The editable source file.
  • A delivery asset such as a GLB.
  • A thumbnail and specification notes.

Start with the shape

3D modeling means creating a shape by placing vertices, faces and other elements in space. Start by combining boxes and spheres to establish size and proportions. This stage is called blockout. Adding fine patterns first does not solve problems in the overall shape or silhouette.

The surface structure serves a purpose

Topology is how vertices and faces connect. It affects bending joints, the appearance of shadows and ease of editing. UVs are coordinates that map a 3D surface to a 2D texture. A rig is a system of bones, control points and other mechanisms for movement; skinning assigns how bone movement affects the surface. A static object that only rotates does not need a complex character rig.

Inspect the results of AI 3D generation

A single image does not show the back or interior of an object. Methods that propose 3D from an image must also estimate unseen shapes. Even when a preview looks good, inspect the underside, holes, thickness, texture seams and editability. Producing a model and making that model usable for its intended purpose are separate stages.

Separate working files from delivery files

Save the editable source separately from the GLB or other asset delivered to the Web. glTF/GLB is used to deliver 3D assets to a runtime, but every feature of the authoring software may not transfer unchanged. After export, load the asset in another viewer or the actual Three.js page and inspect its materials, size, orientation and animation.

Quality checklist

  • Check for holes and overlapping surfaces, including the back and underside.
  • Units, origin, orientation and scale match the intended specification.
  • Required textures and animations are included.
  • Check loading time and frame time on actual target devices.

Diagnose failures

Symptom Likely cause Next step
The front looks good, but the model breaks down when rotated. Acceptance was judged from only one direction. Inspect the full rotation with a turntable view.
Materials change after export. Unsupported nodes or settings. Replace them with supported materials and consider baking the required appearance.

Exercise: Block out a mascot in 3D

Status: not run (not-run). This is a planned exercise, not a measured result.

Task: Build a mascot using only spheres and boxes, then record front, side and back views. Before adding details, judge whether the silhouette still identifies the same character.

Deliverable: A rough 3D model and a comparison of three viewing directions.

Completion criteria: Explain how the information required by an image differs from the information required by a model.

Prerequisite chapters: Map the work and production process · Design a consistent character

  • Three-view character turnaround (Catalog candidate: anime-reference-sheet)
  • Friendly rounded 3D mascot (Catalog candidate: mascot-soft-3d)

Use the experiment worksheet for your plan and evidence

Open the experiment notebook

Primary sources

Source scope checked on 2026-10-04

GLTFLoader loads glTF 2.0 and lists the extensions it supports. Use that explicit list when deciding whether an exported feature can reach the target runtime; do not infer support from a successful file load alone.

MENTAL MODEL / REASONING ORDER

Change one thing at a time.

Intent

Decide what the image communicates, its medium and dimensions, and the meaning you want to preserve.

Sources

Publication dates belong to the source; access dates record when it was checked. Community observations are separate from official statements.

01
Official documentationKhronos glTF ↗www.khronos.orgPublished: Unknown · Accessed: 2026-10-04
02
Official documentationThree.js GLTFLoader ↗threejs.orgPublished: Unknown · Accessed: 2026-10-04
03
Official documentationThree.js Mesh ↗threejs.orgPublished: Unknown · Accessed: 2026-10-04
Saved in this browser only.