OpenAI CLIP Guided Diffusion - Make images from text!

Feel like controlling a powerful artificial intelligence to create images? Then OpenAI CLIP with OpenAI Guided Diffusion is going to be right up your street. Just enter a text prompt, sit back and relax as masterful art appears before your very eyes. Most of the time...

GitHub:

[Dietcheesey]

"I'm just gonna turn that off now cause it's cursed." LOL, you got a subscriber for that.😃

[robertx]

Sorry I'm late, Lord Nerdy 'master of special effects'

[DontBeAJinx]

Why am i getting an error when trying to create video using VQgan+clips on colab?

[Dietcheesey]

Have you tried using colab instead, with a local runtime?

[luisarandas]

Is there a pretrained model for 1024x1024? Thanks

[WillakaPlaeground]

what is the nrupdated cut method and why cant i find out anywhere how to enable it?

[Dsuranix]

there is a better method which you might try, it's fairly accurate, and far faster using this particular model, essentially, you pare down the timestep_respacing (specifically this is set on to under 100, or even sub 50) and run it with an adjusted timeskip in relation to the total, so if 60 respaced timesteps, then 30 would be a "middle ground" for the timeskip option. anyway, it produces extremely high quality images in a fraction of the time, and if you watch progression images, it looks a bit like a slot machine, shuffling through images before landing on one. adjusting the timeskip from that point on can also have extremely interesting results, even by 1 integer at a time.

[eranfeit]

Hi,
Thank you . I have a question .
You mentioned 10 Giga VRAM for the 256 image.

Is it related for the machine ram or the GPU ram ?

Eran

[tristotech]

Wow, cool! I hope with GPT Neo or GPT 3 perhaps this AI will draw better and smoother!

[jasfx]

Bro please explain how to Generate Ai questions from google colab please