Developing sprite sheets with gpt-image-2

aprendendo.next: > Use Codex to split and process frames one by one. This reduces manual effort, though errors can still occur. Have you tried giving it one image with the main frame in the middle (way bigger) surrounded by smaller sprites clockwise (starting at 10:10 am lol) and skip splitting during the normalization? I would try scripting to build “normalization frames” from individual sprites, run them in bulk (parallel) for normalization, then back to scripting to extract only the main frame or normalized samples, and scripting again to stich all together in sprites. But take it just as another opinion, I don’t have all of your context…