Vertex Lab Demonstrates 3D Modeling Workflow Combining V2Fun and GPT-6 Astra

Excavator and character demos show how AI-generated components can be assembled into editable 3D assets

Shanghai, Sept. 11, 2026 (GLOBE NEWSWIRE) -- Vertex Lab, the company behind AI 3D platform V2Fun, has demonstrated a component-based modeling workflow that combines V2Fun’s 3D generation technology with GPT-6 Astra’s planning and coding capabilities. The workflow turns reference images into assembled 3D assets, with individual components retained for further editing. Demonstrations include a detailed excavator and an armored character.

For 3D creators, the approach addresses a practical challenge: generating a detailed object while keeping its parts manageable. By creating major components separately and then assembling them, the workflow produces a scene in which those parts can be inspected, repositioned and developed further. It brings AI generation and code-driven modeling into the same process, with each method used where it is most useful.

GPT-6 Astra interprets the reference image, identifies the main parts and plans how they fit together. V2Fun generates the complex 3D components, including shapes that are difficult to reproduce with simple geometric rules. GPT-6 Astra then uses Three.js, a 3D graphics library, to assemble the returned assets and add regular details through code.

From machine components to a complete character

In the excavator demo, V2Fun generates major components such as the cab, tracks, arm and bucket. GPT-6 Astra adjusts their scale, rotation and position to bring them together. It then adds hydraulic hoses, pivot shafts and cabin controls with Three.js. The finished model combines the complex surfaces of generated assets with smaller details whose dimensions and placement can be controlled directly in code.

The company also presents a side-by-side excavator comparison in its technical blog, showing a build created with procedural code and a hybrid build using V2Fun-generated components. The example illustrates how the choice of modeling method changes the resulting forms and detail. The workflow lets an agent retain procedural control while calling on a specialized model for more complex geometry.

A second demonstration applies the same process to an armored king. V2Fun generates the character’s main assets, including the head, armored body, sword and shield. GPT-6 Astra aligns their proportions and positions, organizes the assembled scene and assesses where polygon reduction is needed. This extends the workflow to faces, decorative armor and other forms that are harder to describe through procedural rules.

Across both examples, the focus is on the steps that connect generation to a usable scene: choosing component boundaries, matching scale, arranging parts and making local adjustments. Keeping those parts separate gives creators a practical basis for subsequent revisions. Further preparation depends on the intended use of the asset, whether that involves animation, interactive content or another 3D workflow.

A specialized 3D capability for AI agents

The demonstrations are part of Vertex Lab’s exploration of V2Fun as a specialized 3D capability within agent-led creative workflows. Delegating complex surfaces to V2Fun allows GPT-6 Astra to focus on planning, assembly and targeted changes, while retaining code for structures that benefit from precise, repeatable control. The work also informs the company’s development of the upcoming V2Fun 2.0.

The full technical walkthrough, including the modeling approach and demonstration visuals, is available in V2Fun’s technical blog.

About Vertex Lab and V2Fun

Vertex Lab develops AI technology for 3D generation and spatial intelligence. Its V2Fun platform brings this work to creators through tools for producing and developing 3D content. The company’s research focuses on complex geometry and consistent textures, with mobile and web experiences supporting different creative needs.

Learn more at v2fun.ai.

Tammy Tan
Tammy@vertexlab.ai