Some #AIart Dwarf Paladin concept art based on the character I just started in #WoW
#AIart #stablediffusion #StableDiffusionXL #sdxl #aihorde #artificialintelligence #artist #art #aiartist #mastoart #fediart #worldofwarcraft #wowclassic #Videogames #gaming #PCgaming
#aiart #wow #stablediffusion #stablediffusionxl #sdxl #aihorde #artificialintelligence #artist #art #aiartist #mastoart #fediart #worldofwarcraft #wowclassic #videogames #gaming #pcgaming
Teilweise erzeugt #stableDiffusionXL erstaunlich coole Bilder 😸
Wie beim letzten Bild der Hut mit der Torte zusammenpasst 👍
Und mit grumpy cat gleich das Wetter düster wird 😅
#AIArt Zoom "Flood"
#ComfyUI #StableDiffusionXL #SDXL
#aiart #comfyui #stablediffusionxl #sdxl
Bueno ya tiene tutorial de #StableDiffusionXL para lo que les guste https://www.youtube.com/watch?v=_ZsZ05Fa8yU
Ok, just found that old issue on "Open"AI's dalle-2-preview repo (https://github.com/openai/dalle-2-preview/issues/6) and with 3.5 billion the SDXL 0.9 base model has apparently as many parameters as DALLE-2 meaning a model the size of DALLE-2 (!) can now run locally on 6GB VRAM.
#stablediffusionxl #stablediffusion
So to conclude: you can now generate images at a 1080p resolution using a two step process involving a 3.5 billion and a 3.1 billion parameter model running locally on a system with 6GB VRAM and to add more context: the first SD versions only have around 890 million parameters.
#stablediffusionxl #stablediffusion
Welp, due to certain recent events in the #StableDiffusion community I was now also able to locally testrun #StableDiffusionXL 0.9 on my Linux-system with 6GB VRAM and I can confirm both the base and the refiner model actually run on my system.
Also: I don't know how ComfyUI (the UI I used to testrun these models as vladdiffusion doesn't support them yet) did that but with it I was actually able to generate images with a resolution of 1920x1080px without running out of memory.
#stablediffusionxl #stablediffusion
RT @hardmaru
“Anime about Linus Torvalds programming Linux”