>>2833000https://files.catbox.moe/rf4bui.pngthe base images could fit some definition of one go i guess. specified characters and simple poses without much overlap do work ok if you separate the prompts with regional prompter.
the fire emblem one was with loras and the pokegirl one was without. you need to use latent generation mode in regional prompter to have it work with multiple loras.
>base gen with regonal prompter with a 1,1,1 split (a generic previous gen or head/torso doodles for controlnet lineart would make it easier to align the area to the the characters but i bruteforced it)>draw a more accurate separation mask in regional prompter with minimal character overlap for upscaling>upscale to 1536x1024>fix possible errors in photoshop>upscale to 1920x1280 (controlnet tile enabled)>fix possible errors in photoshop>disable regional prompter and inpaint each character separately with their corresponding prompt at higher res for final touches and extra detail (controlnet tile enabled)