Welcome to the "Open Diffusion" area at Hugging Face!
We are a small group of volunteers who hang out on discord, and plan interesting things that "normal people" can do in the AI space with consumer hardware. (And keep the results 100% Open Source!)
One of our current projects is "XLSD", a merge of the SD1.5 diffusion model with the SDXL vae, to see how far we can improve the quality with a better VAE. We chose SD1.5, because it is small enough that ONE PERSON can do significant tuning on it at home with a single 4090 card.