Raw LLM Responses

​@DarthBiomech I guess it can feel like a slot machine. I call trying to find good outputs from SD mining, especially when you figure out how much more important the seed seems to be in all my testing*. Once you find a seed that worked well for a prompt it'll work well for similar themed prompts. So I started collecting seeds and their corresponding SD version and dimensions in categories like "landscape seeds", "single figure standing seeds". They are very reliable. It's more like the library of babel in a sense I think, every piece already exists in there, we're just finding them, not creating them. At best these things are the craftsmen/artists someone hired and tasked with making an art piece per their description, sometimes the result is acceptable, sometimes a let down. That's not making art yourself imho.. I'll never take credit for anything AI I might use in anything I make anyway. *Which took me quite a while to go through seed+prompt combos for every step count from 1 to 30 to see how it works, get the inkling seed is much more deterministic for the outcome than prompt. Then set up a testing method to determine if a seed can be bad regardless of prompt (yes) and if seeds are reliable predictors of outcome with changing prompts in the same vein (yes very) (e.g. testing a seed seemingly good for posing with a long object with 30 different variations of object held and changing humans/genders/anthropomorphic animals. And doing that for a lot of seeds that work for other particular things.)