echevarian420, to StableDiffusion

#stablediffusion #aiartcommunity
#aiartwork

#SDXL does seem to have a much clearer association between people and objects, although it doesn't in any real sense understand physics.

The problem will continue to be how to train a model that has no physical presence in the world how to have a concept of just the basics of how things function.

Embodiment is a real issue.

echevarian420, to StableDiffusion

#stablediffusion #aiartcommunity

Eras of Stable Diffusion:

1.5: "We're preserving the entire visual history of humanity in a single model."

2.0: "Please don't sue us, and stop making porn!"

XL: "They already have sued us, nudity is fine, who's scared of Disney?"

echevarian420,

#SDXL seems to understand hands holding tools (like guns) on a level that I don't think even Midjourney has gotten to yet.

That being said this 0.9 version still has a real issue with hands.

echevarian420, to StableDiffusion

#stablediffusion #aiartcommunity #aiartwork

Continued #sdxl testing, I can confirm that it has modern day cultural figures in the dataset.

It also seems to recognize artists who have passed away, living artists seem to have been been removed from the tags the same as in the 2.0 era models.

Not a surprise, just confirming.

echevarian, to StableDiffusion
@echevarian@genart.social avatar

#StableDiffusion #AIArtCommunity #AIArt

I'm cautiously testing how far outside of the 1024x1024 size I can push #sdxl before it starts to get incoherent.

Abstract architecture is probably not the best test case.

ronald, to random
@ronald@ohai.social avatar

next up will be #SDXL 0.9 or whatever latest version, but first experiment with this a bit more while that gets released

ronald, to StableDiffusion
@ronald@ohai.social avatar

toying with running #stablediffusion #ai on my own pc. results are pretty good, considering it's all coming from my own computer and no expensive cloud stuff is involved. used this guide: https://www.howtogeek.com/832491/how-to-run-stable-diffusion-locally-with-a-gui-on-windows/

ronald,
@ronald@ohai.social avatar

@exotaur thanks, ill keep that in mind. the 2.1 is not that old is it? is installing #sdxl also just a matter of dropping a downloaded checkpoint in the same directory?

echevarian420, to StableDiffusion

#stablediffusion #aiartcommunity #aiartwork

I was a little worried that #sdxl wouldn't be able to bring the weird since it seemed to have so much photoreal training material.

echevarian, to StableDiffusion
@echevarian@genart.social avatar

https://www.tumblr.com/echevarian/722413381166956544/testing-out-stable-diffusion-xl-in-the-comfy-ui?source=share

I can confirm that #sdxl is capable of creating NSFW results, as opposed to the hamfisted censorship of the 2.0 era models..

#StableDiffusion #AIArtCommunity

echevarian420, to random

Taking #SDXL for a spin here.

Running locally in Comfy UI..

makeitrad, to StableDiffusion

Day 2 playing with #SDXL #stablediffusion
Houston we have lift off!!! 🚀🚀🔥

image/png

makeitrad, to StableDiffusion

My first #stablediffusion #sdxl local renders 📸

image/png

  • All
  • Subscribed
  • Moderated
  • Favorites
  • JUstTest
  • ngwrru68w68
  • everett
  • InstantRegret
  • magazineikmin
  • thenastyranch
  • rosin
  • GTA5RPClips
  • Durango
  • Youngstown
  • slotface
  • khanakhh
  • kavyap
  • DreamBathrooms
  • provamag3
  • ethstaker
  • osvaldo12
  • tester
  • cubers
  • cisconetworking
  • mdbf
  • tacticalgear
  • modclub
  • Leos
  • anitta
  • normalnudes
  • megavids
  • lostlight
  • All magazines