>>106793650
you can use SVG. chatGPT is quite good at making them, and all LLMs can work with them generally since they are html code and can contain things like comments to describe things in the design.
for example i use it to make a detailed base, then add flair layers that will provide the 'animation'.
otherwise you might be able to get it to create them with javascript grids/arrays
const W = 1; // Wall
const F = 0; // Floor
const G = 2; // Grass
const C = 4; // Flowers (C for cute)
const grid: number[][] = [
[W, W, W, W, W, W, W, W, W, W, W, W],
[W, G, G, G, G, C, C, G, G, G, G, W],
[W, G, G, C, C, C, C, C, C, G, G, W],
[W, G, C, C, C, G, G, C, C, C, G, W],
[W, G, G, C, G, G, G, G, C, G, G, W],
[F, G, G, G, G, G, G, G, G, G, G, W],
[W, G, G, C, G, G, G, G, C, G, G, W],
[W, G, C, C, C, G, G, C, C, C, G, W],
[W, G, G, C, C, C, C, C, C, G, G, W],
[W, G, G, G, G, C, C, G, G, G, G, W],
[W, W, W, W, W, W, W, W, W, W, W, W]
];
then youd convert that grid code into an image with canvas or something.