Case library
Working production cases with sources, tool stacks, evidence and human review requirements.
Generated 3D props for level blockouts
Level designers use image-to-3D tools to make placeholder props quickly, giving blockout levels a spatial feel closer to the final game.
Human reviewPolygon counts and proportions vary; standardise the scale before import.
An agent checks design documents for logic gaps
Before a review, a systems designer has an agent read the design document against a checklist, flagging undefined values, missing states and conflicts with existing systems.
Human reviewThe agent does not know design intent agreed verbally, so question-level flags are often false positives.
An MCP agent checks prefab names and references
Technical artists connect Claude Code to Unity MCP to scan new prefabs each night and report naming violations and missing references.
Human reviewMaintain a manual allowlist for exceptions, such as outsourced assets and third-party plugins.
A repeatable ComfyUI pipeline for item icons
The art team locks item-icon styling into a ComfyUI workflow. Artists supply line art, substantially reducing per-icon time during production.
Human reviewCheck each batch for consistent brand and rarity colours.
Automated localisation terminology checks
The localisation team uses an LLM to check multilingual text against a glossary, catching inconsistent terms and missing placeholders before submission.
Human reviewRewrites involving cultural context require human judgement.
AI placeholder voices for NPC prototypes
During prototyping, speech synthesis provides temporary NPC voices to test narrative pacing. Human voice performances replace them for release.
Human reviewLegal review must confirm voice licensing terms and whether the voices may be used in the final release.
No cases match these filters. Clear the filters or choose different criteria. View all