DALL•E Mini/craiyon - Description-based AI Image generator (that won't let you enter naughty things)

  • 🏰 The Fediverse is up. If you know, you know.
  • Want to keep track of this thread?
    Accounts can bookmark posts, watch threads for updates, and jump back to where you stopped reading.
    Create account
Null said:
Just got it working on AMD. It takes about 3 minutes to generate this though, at 10s/it. I'm using a very large model cpak though, I don't know if that hinders it. Nvidia apparently renders in 4s - oof.
Click to expand...
Oh damn that's really cool.
 
craiyon_164038_Flying_Toilets.png
 
Least Concern said:
Why the Toledo alleyway part?
Click to expand...
Oh, please forgive my odd methods. I thought establishing the setting might help me to capture the American stop sign in its natural habitat. I chose Toledo only because of unconscious interference from Jamie Farr. Of course any other uniquely American location should work similarly.

Least Concern said:
Does it not work with just the stop sign part - you need to give it a place and time?
Click to expand...
I don't know if it's essential, but after a few hours of adjustments it seemed to me that a thorough description of a "real scene" produced more accurate signage.
 
I started using NAIFU instead of the webui, which is basically the NovelAI stable diffusion + backend + frontend. Although it may have been just me being a brainlet with some of the stable diffusion gui settings . My GTX 1070 takes roughly 20-30s per image. It isn't bad, but it might be the push I need to finally upgrade it.
unknown.png

00116-2613152267-black cat, (((((red eyes))))).png
 
You must log in or register to reply here.
Back
Top Bottom