Raw LLM Responses

​@galdoug8918 Yes, I would agree that (the caricaturized stereotype of) prompting big anime tiddies and spamming generate doesn't make generative AI a very useful tool yet, nor garner much artistic merit for that matter. While you might get something passable as the output, problem is you have next to no control over how the end result will look like. You can't really claim authorship over something you had minimal input in creating. Experienced prompter could gain a _higher_ level of control with prompting alone, but still mostly just at a conceptual level. It's not easy to get the exact compositon and detail you're imagining with just prompting, especially in one shot. At a very basic level you could keep iterating with no other tools than prompting, image2image and inpainting, and slowly work your way towards something more resembling what you were thinking. It's a imprecise, and very brute force approach to getting towards a more intentional outcome, but that's where people usually start and many stay. And that's fine, I would still rather see a more polished piece, even if it's more "discovered" than handcrafted, than the raw generation spam we see way too much online. But that's just the start. How much control do you want? Inpaint sketch can give you a very direct way to alter the piece. Familiarity with your models and loras can help you nail down your desired style - or you might want to use controlnet reference, or the nuclear option, train your own loras. You can use any number of controlnets for transfering over existing outlines, or to use a crude sketch as your basis, or extract outlines from other sources like photos. Or, you could make a crude composite in photoshop and use that as your starting point if that's something you're more familiar with. You could work with 3D reference with depth maps. If you need very fine control over posing your characters, you can pose openpose bones. Need an easy way to block out your composition and colors? T2ia got you covered, and if you need crazier control, regional prompting, segmentation maps, etc. Honestly with just the basic tools and a little creativity there is no excuse why you couldn't reach your creative vision with some effort. Any method that's intuitive to you, or fits the project best could probably be leveraged one way or another. But that takes some familiarity with the tools and techniques, or dare I say, skill in the craft. The limit of how much agency you can to have over the process is just how much effort you _want_ to put into it, and the fact that you _can_ leave some aspects to chance could be consireded a unique advantage of its own. And I know that's a lot of jargon that means nothing if you're not familiar, but that's almost the point. I believe this misrepresentation that generative AI couldn't be a viable _tool_ in creative processes is often coming from a place of ignorance more than anything else, but while that's at least something I can empathize with, it definitely doesn't make me disagree with that take any less. I don't mind if someone doesn't like ai art for any number of subjective reasons, that's fair. But for me it is still already a damn strong tool, and one that has such a distinct approach that I would consider it its own artform entirely.