Model was trained on 512x512 images, so it works best on this resolution.
Model was trained on D&D monsters, heroes and master characters (NPCs). So it works best for medieval fantasy setting. It's recommended to use trigger words, that define monster type.
But it's not overtrained only on this setting, as for me works pretty good with almost everything I tried. You can not use trigger words at all, that would be fine too, sometimes that's even better.
Also you can try mix them to get interesting results. Check all example images for inspiration.
Link to huggingface: https://huggingface.co/Zapper/top-down-token-v1-0