People asked me to make a twitter account. Not sure what/how much I'll be posting here, probably just updates and maybe one-off AI art posts. At least I can use the API key to get my RSS feeds working again.

Ended up fixing up this gen a bit while my GPU was busy preprocessing Danbooru2021 for a longer training run. #AIArtwork #aigenerated

I guess I should post here sometimes...

Random failed #StableDiffusion training run. Using segmentation maps seems to have left the model with little to no idea on what to do with the background, making it look like a photo with a cheap filter over it. #AIart

No external editor used, all composited inside ComfyUI #AIart

#StableDiffusion #AiArt

Network distributing #StableDiffusion via ComfyUI. github.com/city96/ComfyUI…

Introducing, a 17B proof of concept self-merge of Flux 1 Dev 12B. By applying questionable LLM merging techniques to image models, I've created this slightly cursed, larger version of Flux. It's very undertrained but surprisingly coherent even like this huggingface.co/city96/Flux.1-…