Quinn I’ve tackled that exact trade‑off before. Start with a clean topology: keep the main silhouette and window lines, then strip out anything that won’t be seen from the camera’s typical angles. Use a low‑poly base mesh for the whole block, then apply normal or parallax maps to give the illusion of depth without extra vertices. If you need some buildings to stand out, bump them up with a single higher‑detail LOD that you swap in when the camera gets close. That keeps the overall draw‑call count low but still lets each structure feel unique. The key is to test on the target hardware early, so you’re not surprised when you hit a frame‑rate wall. Keep the workflow repeatable, and you’ll maintain a healthy balance between fidelity and performance.

Quinn You’re right, the edges carry the story. Start each building with a clean silhouette loop, then prune everything that doesn’t contribute to that shape. Keep the quads aligned, avoid weird splits that create noise in the render. Once the base shape is solid, add a normal map to carve in those subtle creases and detail. That way the mesh stays light, the form stays true, and the overall feel isn’t cluttered. It’s the same discipline you use in any planning—set the framework, then layer only what adds real value.