did you get this phrase from my fav borges short story, or come up with it coincidentally?
I shamelessly pilfered “unanimous night” for a short story I wrote a few months ago for a zine that’ll come out eventually
did you get this phrase from my fav borges short story, or come up with it coincidentally?
I shamelessly pilfered “unanimous night” for a short story I wrote a few months ago for a zine that’ll come out eventually
Exploded diagram of a Cyber-brain
Ever since I started using the VQGAN+CLIP GAN I’ve been thinking to myself that I should use something like this to create the basis for some assets to make some kind of horrible game. Maybe something on the GZDoom Engine. The ideas haven’t quite cohered enough to give me any sense of direction on that point.
Let’s try some Shmup-related prompts, eh?
R-Type Gothic
Fantasy Zone for NES painted by Salvador Dali
Gradius for Super Nintendo with RTX on
Parodius for Sega Saturn drawn by Leonardo DaVinci
Keio Flying Squadron for Sega CD painted by Francisco Goya
I feel as though some of these could be good games in their own right. It definitely seems to understand R-Type better than it does some of the other games.
oh yeah no i absolutely got it from that story. simply my favourite sentence
my only hobby is making nightmare collages now
CHOOSE YOUR FIGHTER
gay grinch dressed in leather

i love how it even gave him a little christmas tree
im always intrigued by what the ai thinks of things, like is the grinch the green man? is it who he is or what he does, the man or the deed
Who shot JFK saturday morning cartoons

Super Mario eating pizza in the big city

A videogame about cats

Anime jrpg video game about cats

Metroid Castlevania 4k 60 fps

Samus Aran and Simon Belmont Fight Dracula

why am i doing this






Last one: Cyberpunk Simpsons Slime

Ooh, the GAN seems to have a solid appreciation for the visual elements of a Bosch painting. I’m gonna have to add a few of those to my own repertoire.
I got good results with Francisco Goya as well, as evinced by
The Teletubbies by Francisco Goya
Are you using a reference image for these or does the ai come up with that on its own? I feel like it’s rare to get stuff with such distinct individual figures.
sometimes the ai gives up the game too easily, bruce campbell is trying to crawl out of that image so hard and the text is all 
doom

evil dead

evangelion

minecraft

Naw, I’m not feeding any references to the one I’m using. However, the VQGAN+CLIP has access to several large databases of images. I’m using the largest of them, which is apparently, “imagenet_16384.” I suspect that’s how it knows so many references to artists, games, movies, television, actors, etc.
I’ve started feeding the output from VQGAN+CLIP into the Deep Dream Generator, despite the fact that I’ve had difficulties getting good results out of the latter.
Here’s one of my latest mashups on the Deep Dream Generator,
“Ultraborne: Vice City”
I might be mistaken but I can’t help feeling as though this could be the basis for a compelling DLC/mod/TC for Ultrakill once the game is complete and (at least theoretically) moddable.
However, the picture itself is noisy and I feel as though it lacks the refinements I see from the popular posts on the Deep Dream Generator. I’m attempting to determine why this is.
I believe I’m beginning to understand the sort of cycle that I need to undertake in order to tease good results with less noise out of the Deep Dream Generator. I definitely recommend reading the manual on that one.
I also seem to have forgotten the URL for at least one AI I was playing with a few weeks ago, which is a shame but maybe I’ll remember it if I sift through my recent Twitter follows (not a recommended course of action, btw). However, I can list at least three tools I’m working with right now.
Tools
Since the VQGAN+CLIP is fairly limited in terms of the resolutions it is willing to output (at least in the absence of a Collab Pro subscription) so my thought is that perhaps the upscaler can be used on output from the former to provide the level of detail needed to reduce the noise that results from feeing VQGAN+CLIP output into the Deep Dream Generator.
I won’t know whether or not this theory holds any water until I’ve been able to test it. Sadly, the Deep Dream Generator is stupendously miserly with its server time.
Feeding the noisy output of the DDG into the upscaler does not seem to beget good results, so we cannot necessarily approach this process transitively with respects to optimizing the pipeline for the tools that give us the most computing time. So it goes. :[
Also, I’m not terribly enamored with upscaler.ai (which I may have said already) because its interface acts extremely peculiar on both my Firefox (possibly a problem with extensions like uBlock Origin) and Brave (probably just Brave doing some esoteric Brave thing). However, it doesn’t seem to have any immediately evident usage limits which is nice.
my life in the bush of ghosts …