After some months making my own objects and plugins for use in Perchance, I compiled, from googling around and finding a list of 800ish artists that are recognized by SD1.5, this list in to a format Perchance and Javascript can recognize. https://perchance.org/artistslist#edit
From there, with the plugins (javascript objects) I had created, I made https://perchance.org/loveandlight in largely javascript but with minor Perchance Script and using Perchance’s ai image plugin, which is a wonderful thing shared freely by the creator of Perchance. I believe in winning lategame. So love and light was built to sift the 800ish artists in to useable format. Prompt is created in the right button, the first two images can have custom additional prompts, and then the other 4 images incorporate primitive ‘machine learning’. Essentially, the user responds likes and dislikes to the images generated by specific artists. Then, especially the last image, incorporates artists in ratio according to like amount. I have been generating some on Love and Light most every night and have maybe 90% of the artists at least weeded for bad ones. In theory, I will eventually hit 100% and it wont stop there as the exact ratios will become more and more tuned. Love and Light starts looking bad since no refinement has happened (if you try it), and takes a long time to start generating good content. I then paste over to my own list of artistlists the artist lists I like every night. As this list grows, it becomes a store of better and better lists of more tuned artists that ‘collab on my pictures’.
beautiful-people then pulls from these lists of artists, compiled via love and light, to back it’s ‘styles’, which i believe makes it unique among ai art generators i have seen. I have each style in beautiful-people include occasionally new artistgroupings from the artistlist and can personally choose whether to add them to each style. So beautiful-people, love and light, and the list between them are all constantly growing better with waaaaay late endgames as far as how they are constructed. I just got the anime style decent lastnight and the steampunk also. I learned node.js servers a while ago and am making my own version of the gallery starting today.
Perchance uses models mostly chosen by the maker of perchance, which i am happy with, so that is what i am working with here. will get in to sd3 when my computer comes and that will be entirely separate but possibly use what i have learned from all this.
currently have to manually upscale
so as to what i have made: largely something in javascript and perchancescript within perchance that has been many layers of creation over months and we are seeing finally start looking beautiful with the surface layer being https://perchance.org/beautiful-people now that I have made it through the majority of the artist list via love and light. there should be styles here unseen elsewhere, due to the artists, with decent variation within each button. Just became good enough I feel it shareable and am proud of it. :) There is no profit to me and i will be paying for the server for the gallery myself, just as the maker of perchance pays for the servers for the ai imaging themself. Moral purity matters.
https://sh.itjust.works/post/16191939
and here is the real post in which I explore this subject.
Thank you!
deleted by creator
After some months making my own objects and plugins for use in Perchance, I compiled, from googling around and finding a list of 800ish artists that are recognized by SD1.5, this list in to a format Perchance and Javascript can recognize. https://perchance.org/artistslist#edit
From there, with the plugins (javascript objects) I had created, I made https://perchance.org/loveandlight in largely javascript but with minor Perchance Script and using Perchance’s ai image plugin, which is a wonderful thing shared freely by the creator of Perchance. I believe in winning lategame. So love and light was built to sift the 800ish artists in to useable format. Prompt is created in the right button, the first two images can have custom additional prompts, and then the other 4 images incorporate primitive ‘machine learning’. Essentially, the user responds likes and dislikes to the images generated by specific artists. Then, especially the last image, incorporates artists in ratio according to like amount. I have been generating some on Love and Light most every night and have maybe 90% of the artists at least weeded for bad ones. In theory, I will eventually hit 100% and it wont stop there as the exact ratios will become more and more tuned. Love and Light starts looking bad since no refinement has happened (if you try it), and takes a long time to start generating good content. I then paste over to my own list of artistlists the artist lists I like every night. As this list grows, it becomes a store of better and better lists of more tuned artists that ‘collab on my pictures’.
beautiful-people then pulls from these lists of artists, compiled via love and light, to back it’s ‘styles’, which i believe makes it unique among ai art generators i have seen. I have each style in beautiful-people include occasionally new artistgroupings from the artistlist and can personally choose whether to add them to each style. So beautiful-people, love and light, and the list between them are all constantly growing better with waaaaay late endgames as far as how they are constructed. I just got the anime style decent lastnight and the steampunk also. I learned node.js servers a while ago and am making my own version of the gallery starting today.
Perchance uses models mostly chosen by the maker of perchance, which i am happy with, so that is what i am working with here. will get in to sd3 when my computer comes and that will be entirely separate but possibly use what i have learned from all this.
currently have to manually upscale
so as to what i have made: largely something in javascript and perchancescript within perchance that has been many layers of creation over months and we are seeing finally start looking beautiful with the surface layer being https://perchance.org/beautiful-people now that I have made it through the majority of the artist list via love and light. there should be styles here unseen elsewhere, due to the artists, with decent variation within each button. Just became good enough I feel it shareable and am proud of it. :) There is no profit to me and i will be paying for the server for the gallery myself, just as the maker of perchance pays for the servers for the ai imaging themself. Moral purity matters.