The Neon Harbor titles gave me five weeks for a full CG port city. I spent the first weekend on grey boxes — and it was the most important weekend of the project.
Start with the camera, not the buildings
Big environments tempt you to model first and frame later. That's backwards. I blocked the entire 75-second camera move on day one — one spline, eleven beat markers for the cast cards — and only then asked what the camera actually sees. Two-thirds of the "city" never appears on screen. It was never built.
The blockout was boxes with a strict rule: every box carries its final name (dock_04, crane_mid_02) from the first hour. Renaming week-three geometry is where schedules go to die.
One reference board to rule it out
Forty images, no more: night ports, sodium vs LED signage, wet asphalt at three rain levels, ferry wakes. When a choice wasn't on the board, it didn't go in the frame. Constraints are a render farm for decisions.
Fog cards before volumetrics
The depth in the final sequence reads as expensive volumetrics. It started as thirty gradient cards. Cards render in seconds, art-direct in minutes, and teach you exactly where the real volume needs to go — which turned out to be three shots out of eleven. The rest stayed cards, and nobody has ever noticed.
What I'd do differently
I'd previs the water on day one instead of day four. Reflections sell half the sequence, and the ferry wake took two iterations I hadn't scheduled. Next city build gets a water test before a single building.
The blockout weekend closed with a watchable grey animatic, approved camera path and a shot list with real polygon budgets. Weeks two to five were craft, not discovery — which is the whole point.