This is relaunch of "Darkdream Diffusion". General purpose model trained solely on my art and further adding to dataset its generations. I trained relaunched version from ground up with all the experience i gathered working on "Darkdream Diffusion", so it's improved in every possible aspect.
It is reccommended to run without VAE.
RFKTR's darkdream is free of any shady licenses and additional fees and it will stay that way - which means i won't come after you if you use it for your comic books, art books and so on. I'll be happy to see what you will make with it. It's purpose is to be free and dependable model for all the community, if you see it on generation services that charge you to use it - know that they're acting against my will.
From now on updates to it will be much more few and far between.
Description
This version was trained on whole new vanilla+ model that i created specifically for stable training.
improvements to dramatic lighting
better minimalism
better watercolor
better abstract
more stability
it works great with both - clip skip 1 and clip skip 2 now
no longer needs trigger word
works with both anime and non-anime VAE's, use whatever you like!