I made Sydney Pixels, a zoomable isometric illustration of inner Sydney, from the Harbour Bridge to Central. I got the idea from Isometric NYC.
I captured the city from Google’s photorealistic 3D tiles at a fixed camera angle, then illustrated it in pieces. GPT helped set the style and drew 204 tiles. I ran FLUX.2 klein locally on an M2 Max for the other 38. On my setup, three inference steps took about 50 seconds per local tile.
The fiddly part was keeping a road or roof lined up where two tiles met. I used overlapping edges as context, blended the seams, and exported a Deep Zoom pyramid so OpenSeadragon could load the detail as you zoom.
One early version put a second Opera House in the middle of the harbour. For water-heavy tiles, I stopped supplying the general style reference and explicitly told the model to preserve only what was in the source view.
If you’ve mixed image models across a large canvas, I’d be curious how you kept the style consistent across seams.