AI Generates Orthographic Views for 3D Conversion
Fresh-Resolution182 · reddit · 2026-07-17
The author built an interactive 3D showcase site this week featuring real aircraft and rocket hardware, including the F-22, F-35, Starship, and Raptor engines.
The workflow has three steps: First, GPT Image 2 is used to generate orthographic reference images (front/side/back) for each object. Next, these images are fed into Hunyuan 3.1 (within ComfyUI) to create actual 3D geometry rather than flat renders. Finally, based on an open-source template, GPT 5.6 Sol rebuilds the entire site to closely match the layout, typography, and dark visual style of the official SpaceX website.
The author concludes that the hardest part is actually creating the orthographic reference images. As long as this step is clean, the subsequent 3D generation and site reconstruction proceed smoothly.
More from coding & agent
- A Forward Deployed Engineer job really has three stages: audit, evals, deploy — blaizedsouza · 2026-07-22
- 438 sealed tests show coding agents prefer DIY over third-party databases — cramforce · 2026-07-22
- Building a stock research agent turned out to be a prompt-design problem, not a data problem — jwstrategies · 2026-07-22
- Tenable and AWS launch a Black Hat build event for open-source security agents and MCP servers — Dave_Maynor · 2026-07-22
- Discussion: How Much Permission Would You Give Your Background AI Agents? — Total_Drag7439 · 2026-07-22
- Codex helps build Valdiluce, an open-world game with climbing, gliding and gondolas — Dimillian · 2026-07-22