Build Structured 3D Parts
Build clean, repeatable parts such as shells, brackets, pipes, frames, and mechanical components.
V2Fun AI 3D
Choose the right method for each part — AI-generated meshes, procedural Three.js geometry, or hybrid modeling — then assemble in Three.js and deliver a GLB with editable source.
What You Can Do
Use reference images or an existing 3D model to create a complete 3D asset. Build structured parts, generate organic forms with AI, combine them, and export an editable result.
Create Organic 3D Forms
Use AI to create people, animals, fabric folds, and other detailed shapes that need a natural look.
Combine Parts into Complete Models
Bring generated parts, existing assets, and supporting structures together as one complete 3D model.
Export and Continue Your Work
Export the finished model as a GLB with editable source files, so you can review, refine, or continue it later.
Inputs & Outputs
Inputs
- Reference ImagesOne or more clear views of the object, character, or scene.
- Design BriefSubject, composition, recognizable features, target use, and deliverables.
- Assets and RequirementsReusable GLB or Three.js assets plus optional output constraints.
Outputs
- Assembled GLBA validated binary glTF model, checked by export and clean-scene reload.
- Three.js SourceReproducible JavaScript/TypeScript construction and assembly code.
- Parameters & RecordsDimensions, transforms, part relationships, inspection records, and known limitations.
EnvironmentJavaScript or TypeScript with Three.js for local geometry and export; Python 3.9+ helpers for API calls and status. AI generation uses the V2Fun API and requires a configured V2FUN_API_KEY.
Made with V2Fun AI 3D
Work 01
Shrine Courtyard
Three V2Fun architectural assets combined with a procedural stone terrace, planting, pond, rails, lanterns, and final scene assembly.
View Work →
Work 02
Punk Rocker Character
A V2Fun-generated performer and guitar assembled with procedural microphone hardware.
View Work →
Work 03
Victrola Gramophone
A V2Fun-generated brass horn combined with a procedural wooden cabinet, record player, and record collection into a complete GLB assembly.
View Work →How It Works
- 01
Plan & Prepare References
Identify the subject, important features, part boundaries, and modeling methods. Prepare required reference views and define a finite API call plan based on the approved workflow.
- 02
Build Ready Parts
Construct regular geometry locally and generate textured AI parts using the approved plan, ensuring each component is ready for assembly.
- 03
Assemble
Align scale, axes, anchors, contact surfaces, and materials in Three.js to combine individual components into a complete 3D asset.
- 04
Validate and Deliver
Inspect the rendered model, export GLB, reload it in a clean scene, and deliver source files with validation records.
Get Started
Turn your reference images or existing model into a complete 3D asset. Follow the steps below to install the skill, check your setup, and start your first project.
Install the Skill
Download the skill from GitHub, follow the setup guide, and explore a ready-to-use example.
View on GitHubAPI access
Using V2Fun remote generation requires an API Key. Eligible accounts receive 100 API credits after successfully creating their first Key, once per account. Local checks and local-only workflows do not require a Key.
Get API Key# Install: extract the skill into ~/.codex/skills/v2fun-ai-3d/
# (or $CODEX_HOME/skills/v2fun-ai-3d/)
python3 scripts/v2fun.py doctor --project examples/local-inventory
python3 scripts/v2fun.py models --project examples/local-inventoryUse $v2fun-ai-3d to rebuild the object in these references.
Choose a method for each part, explain the API call plan,
and deliver a GLB with its Three.js source project.Expected OutputAn assembled, validated GLB, its Three.js source project, parameters, and inspection records.
Requirements & Limitations
Requirements
- Codex with local Skill support, local file access, and script execution.
- Python 3.9+ (standard library) and JavaScript/TypeScript with Three.js plus a WebGL browser for local 3D work.
- A V2Fun account with a configured V2FUN_API_KEY for AI generation; local-only work needs no key.
Limitations
- This is a visual-mesh workflow; exact CAD/STEP output, automatic rigging, certified print preparation, and video production are outside its scope.
- Hidden surfaces are inferred rather than recovered as measured facts.
- Standalone texturing, retopology, and automatic rigging are not implemented; model-driven invocation and live generation are not yet verified.
Frequently Asked Questions
Does the Skill generate every component with AI?
Not necessarily. It selects AI generation, procedural modeling, hybrid workflows, or asset reuse for each component based on its structure and complexity.
Can it create 3D assets from a single reference image?
Yes. The Skill can analyze a single reference image, identify visible features and part structures, and infer unseen surfaces during the modeling process. Additional views can improve reconstruction accuracy when available.
How does the Skill build complete 3D assets?
The Skill follows a structured workflow that includes reference analysis, component planning, part generation, Three.js assembly, GLB export, and final validation in a clean scene.
Does it combine procedural modeling with AI-generated assets?
Yes. The Skill combines procedural geometry with AI-generated components, allowing regular structures to be built locally while complex textured parts are created through AI workflows.
What does the Skill deliver?
The Skill delivers assembled 3D assets with exported GLB files, source components, and validation records to support further use and integration.
Explore More
Start Building with V2Fun AI 3D
Provide a visual reference and let your agent plan, generate, assemble, validate, and prepare a complete 3D deliverable.