The following images were made using my custom build of Disco Diffusion, photo-manipulation, photobashing and sketching through Gimp and Inkscape, and post editing/composition in darktable.
It's slow, it's taking you days to do what others can do much faster on other AI, it's hard to get coherent result, it has its own will, things always turn out very whimsical, why don't you just change program?
1) I can't run Stable locally as my GPU is absolute garbage
2) I don't want to pay for generations, so that excludes premium versions of GoogleCollabs, MJ, etc
3) I want to be able to control every single parameter of my image generation, from the contrast level to the type of noise generated, and to schedule my cuts depending on the piece I have in mind. Only DD gives you that much creative control over things so far, besides locally run SD (which I already said I can't run)
4) I actually really like the surprise-whimsical aspect of things with DD, it gives every result a random twist that I really like.
5) I'm too lazy to have to pick the best image out of a thousand generations. With Disco Diffusion, when it takes more than an hour for 1 quarter slice of the image to generate, you only do one. The settings have to be perfect, the prompt as well, I find it more challenging than simply click and repeat and pick the best out of all.
Plus, let's be honest, I didn't spend all this time learning and tweaking the code of Disco Diffusion, training models that work best with it, to just jump off to other platforms simpler to use and more accessible, with less creative control.
