>>82546785Pekora anon, is that you?
>>82547568Nice peko.
>At least with AOM, if properly tagged, you could use it in negatives to get flatter imagesAOM had AI generated pictures within the training? That's news to me.
>to get flatter imagesI am unlike many who likes 2.5D AOM style, I wouldnt ever want it to go flatter. We have plenty of models like that now but I havent seen another 2.5D hybrid. Then again it's dangerous territory and I don't want it to go into 3DPD range.
Like those Shiori gens above>The part about forgetting previous styles is due to how models are being trained... it is not an adding onto the existing knowledge and it does not add new neurons or parameters, it changes the existing ones.Clearly when they have the priority of overwriting existent artist styles, it says a lot about the model, no? Who the fuck would want xXniggerprompter69Xx ""artstyle"" when I can have waterkuma's?
>Maybe with Liquid Neural Networks we can have something that works this way but they are far in the future if they will even work for imagegenIf a model sucks and is not to my liking, I'll simply not use it when there are much better alternatives, simple as. AI is progressing with each day, so far it has been going in a great direction. But I still cannot fathom why would anyone use AI gens for a dataset when training a model.
>I did comment on your postsOh that was you? My bad. I didn't realize. Disregard my complaints then.
>Hey as long as you get what you want, I'm happy. Yes, I am very satisfied with your model. It is dangerously close, if not identical to what I was getting on NAI3, so thank you again for making it and sharing with me. Picrel is also made with your model with bare minimum tags.
I'll post the comparison test images later today. I need to work on my past gens and upload to pixiv.
