Upgrade your account to a Community Supporter account and remove most of the site ads.

Reply to thread

Message: <blockquote data-quote="Yaarel" data-source="post: 9394668" data-attributes="member: 58172">A difficulty with AI, it seems only able to handle a certain amount of variables at one time. It is difficult to obtain a precise image. After the resulting images finally does get a certain part of the prompt right, an other part of the prompt that was previously fine goes wrong.Does anyone have sense of how many variables the AI can reliably handle? Or is there a way to lock-in a semisuccessful result, and modify it from there? (It might be viable to use a specific image as a starting point, but I am unfamiliar with the tools that do this.)I noticed the problem when creating Norse images. There is so much misinformation about Norse content in popculture, that simply entering the term "viking" inevitably introduces bizarre errors − like weird fur leggings, arm bracers, and ginormous weightlifting belts − nevermind the horned helmets.I find myself needing to reconstruct images from scratch, indirectly, focusing on each element separately. For example, to create a Norse dragon (dreki), it was necessary to avoid the word "dragon". Instead, I blended "european adder" with "leonine head" and "goat horns", with surprisingly good results that kinda did look a historical wood carving of a dreki.But complex reconstructions with many variables to keep in place, it seems too difficult for the AI to keep track of all of the elements. It inevitably introduces nonsense (omissions and misunderstandings) even when earlier it had been getting this part of the prompt correct.</blockquote>

Verification