Build a Steam Rail Cart with GPT-6 Astra and V2Fun

Overview

This case turns a single collectible-style concept into a complete 3D asset combining a seated train driver, a miniature steam rail cart, and a short section of track. The supplied references show the intended bearded character in dark-blue workwear, a green horizontal boiler with brass fittings, red flanged wheels, exposed connecting rods, an open rear seat, and clear contact points around the lever, seat, and foot supports.

GPT-6 Astra interprets the subject hierarchy and coordinates the assembly, while V2Fun handles the organic character form. Regular mechanical elements are represented procedurally. This division follows the broader GPT-6 Astra × V2Fun model-collaboration workflow.

Tools Used

• GPT-6 Astra: reference interpretation, component planning, spatial relationships, and assembly calibration.
• V2Fun: generation of the driver as a textured 3D component. The platform's AI 3D model generator supports reference-led asset creation.
• Three.js / procedural geometry: the structural frame, running gear, body, seat, mechanical equipment, controls, and supports shown in the component study.
• Visual references: a complete concept, an isolated character reference, and an inferred front-side-rear study.
• GLB: the supplied final delivery format; this 3D export-format guide explains when GLB is useful in broader workflows.

Step 1 — Read the Complete Concept

The initial image establishes the target as one self-contained character-and-machine asset rather than a loose collection of props. Its front three-quarter view makes the driver, boiler, wheels, rods, seat, controls, and track base simultaneously readable.

Step 2 — Prepare Character and View References

The driver is isolated from the vehicle so his stocky proportions, beard, soft work cap, seated pose, clothing folds, and hand position can be considered separately. An inferred three-view study supplies front, side, and rear guidance for the combined vehicle. These inferred views clarify otherwise hidden relationships, but they remain interpretations rather than measured source data. For projects that begin with several captured angles, see the multi-view 3D modeling workflow.

Step 3 — Decompose the Asset

The process image divides the build into six visible groups:

1. Character
2. Structural frame and track base
3. Running gear
4. Boiler body and seat
5. Mechanical equipment
6. Controls and supports

This separation gives the agent explicit units to position and revise while preserving important relationships such as the wheel-to-rail alignment and the driver's contact with the seat, lever, and foot supports.

Step 4 — Generate the Organic Component

V2Fun generates the driver as the complex organic component. The evidence shows the character carried into the final assembly with his blue work clothes, cap, beard, seated posture, and one hand placed at the control lever.

Step 5 — Build the Regular Mechanical Parts

The component study labels the remaining major groups as Three.js procedural. Cylinders, discs, rectangular plates, straight rods, repeated sleepers, wheels, pipes, pedals, and supports are suited to this approach because their forms and relationships can be described directly.

Step 6 — Calibrate and Assemble

GPT-6 Astra brings the generated and procedural pieces together, adjusting their relative placement and scale. The final render shows the driver seated behind the boiler, four wheels positioned on two rails, side rods aligned with the running gear, and the controls arranged within reach. The presentation identifies this image as a render reloaded from the delivered model rather than an AI redraw.

Workflow

Concept reference → Isolate character → Infer additional views → Decompose components → Generate character → Build regular machinery → Calibrate contacts and alignment → Assemble → Export GLB

Prompt

Create a complete standalone collectible-style asset featuring a short, stocky adult train driver with a thick beard, soft work cap, and dark-blue overalls. Seat him at the rear of a miniature steam rail cart with one hand gripping a brass control lever. Build the front from a horizontal cylindrical boiler, circular face, and vertical smokestack; use a rectangular chassis, four flanged metal wheels, exposed straight connecting rods, an uncovered rear seat, and a short two-rail base. Use dark-green enamel, brass pipes, and red wheel hubs. Combine realistic proportions with mild stylization, retain clear contact points between the driver, seat, handle, and foot supports, and show the whole subject from a slightly elevated front three-quarter view. Use a white background, soft studio lighting, contact shadows, sharp focus, and refined PBR-style materials. Exclude text, numbers, logos, watermarks, complex scenery, steam occlusion, motion blur, collages, and exploded-part presentation.

Result

The delivered asset is 85-v2-final.glb, a model/gltf-binary file containing the assembled steam rail cart presentation.

• File size: 42,572,840 bytes
• Visible result: one seated driver, one miniature steam vehicle, and a short track base
• Delivery: static GLB assembly

The supplied evidence demonstrates visual assembly and GLB delivery. It does not establish topology quality, UV quality, polygon count, physical dimensions, printability, animation readiness, engine performance, or commercial-use rights.