Codex integrates reviewed vehicle-art variants
Lao Liu reports a Codex art pipeline that turns an approved vehicle design into positioned body and lighting layers, sixteen day-and-night variants, icons and an in-game car.
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Use the approved vehicle design to prepare body and lighting layers. Open a positioning tool against the real background for my visual decisions. After approval, generate consistent variants and integrate them into the engine, checking rotating-wheel geometry.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- Lao Liu
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- Lao Liu reports a Codex art pipeline that turns an approved vehicle design into positioned body and lighting layers, sixteen day-and-night variants, icons and an in-game car.
- Human involvement
- Liu selects direction and approves designs and positions by eye. Agent self-review is fallible, and the new game’s release is not established by this account.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Approve a vehicle design before generating production layers.
- Separate wheels and prepare body, wheel-well and lighting assets.
- Use the positioning tool against the real background to adjust visual parameters.
Lao Liu uses Codex to organise and finish vehicle art for a bottom-of-screen idle driving game. The pipeline begins with a design-confirmation image, then separates wheels, completes wheel wells and creates lighting layers after he approves the design.
The agent opens a purpose-built positioning tool showing the actual driving background and a reference vehicle. Liu adjusts body and wheel placement, light areas and shadows by eye. Approved positions feed the repeatable stage: sixteen day-and-night colour variants, UI icons and engine integration. The reported outcome is a new car appearing in the game without him manually performing every editing and import step.
Wheel geometry exposes an important constraint. Generated wheels can look acceptable while still wobbling during rotation. Liu changes that part of the pipeline to Blender-based circular geometry; a normal map and Godot lighting provide a moving shadow. Generative images and deterministic tools therefore serve different parts of the same production process.
The author retains visual direction and approval, and acknowledges that agent inspection can be wrong. The account shows an actual project workflow, not a benchmark proving every asset passes review or that the new game has shipped. Its useful detail is turning subjective placement into an interactive human checkpoint, then letting the agent resume the stable variant-generation and integration work.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Blog · Lao Liu
Frequently asked questions
Where does human judgment enter?
Design approval and the custom positioning tool, where the creator adjusts bodies, wheels, light areas and shadows against the actual background.
Why use Blender for wheels?
Generated images did not reliably guarantee circular geometry or rotation-compatible lighting.
What followed placement approval?
Sixteen day-and-night colour variants, UI icons and integration into the running game.
More details
Remaining steps
- Generate sixteen approved day-and-night variants and UI icons.
- Integrate assets into Godot and inspect wheel rotation and lighting.
About this account
Codex coordinates generation, local image processing, parameter tools, variant production and engine integration, resuming repeatable work after human visual decisions.
Firsthand account by Lao Liu; original publication date verified against Primary creator byline Published 2026-08-15. Outcomes are the author’s reports, not independently reproduced. Suggested prompt written by Agents In Use; not the author’s original prompt.
Credit: Lao Liu
Source published
Added to Agents In Use
Source checked
Tools and topics
Codex · Image generation · Positioning utility · Blender · Godot
Vehicle-art variants · Human visual positioning · Rotating wheel geometry
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.