# Three-minute judging walkthrough

0:00–0:20 — Explain the problem: creative scene exploration plus a separate game for practicing route planning during hurricane response. Identify SDXL and Waypoint on Hugging Face.

0:20–0:55 — In the local application, enter a short scene description, generate an SDXL HIGH seed and inspect the preview. If live generation fails, explicitly select a saved seed.

0:55–1:25 — Start genuine native720 Waypoint exploration, demonstrate WASD and drag-look, pause/resume and refresh. State that generation and display FPS are measured separately, and the scene can drift.

1:25–2:20 — Play a short Rescue Mission. Explain the goal, hazards and deterministic scoring; inspect stars and A* route comparison. Do not claim the video and grid have the same geometry.

2:20–2:40 — Open AI Showdown and explain that it uses previously evaluated PPO results, not live training or new performance claims.

2:40–3:00 — Open this Hugging Face Space and show model attribution, native720 evidence and limitations. Summarize what was engineered: persistent inference, bounded streaming, correct acknowledgments, lifecycle and reproducible benchmarking.

This is a proposed presenter script, not evidence that a new three-minute walkthrough was performed during packaging.
