There is an AI plugin for krita that lets you define regions, selection bounds, sub-prompts, control nodes, and lots more control over a given image generation model than standard Automattic or comfyUI workflows...down to 'put an arm wearing armor here' for example in my RPG NPC token writing.

It has full image generation mode, it has an animation mode, it has a live mode where you can draw a blob of images and it will refine it 2-50 steps only in that area.

So you are no longer doing per line stroke and saved brush settings, but you are still painting and composing an image yourself, down to a pixel by pixel rate. It's just that the tool it gives is WAY more compute intensive, the AI is sort of rendering a given part of a drawing as you specify as many times as you need.

How much of that workflow is just prompting a one-shot image, vs photoshopping +++ an image together until it meets your exact specifications?

No, the final image cannot be copyrighted under current US law in 2026, but for use in private settings like tabletop RPGs...my production values have gone way up and I didn't need to get a MFA degree in The old Masters drawing or open a drawing studio to get those images.