post pictures your computer has generated ( algorithmic art )

[catalog]

using https://colab.research.google.com/g...in/StyleGAN3+CLIP.ipynb#scrollTo=V_rq-N2m0Tlb

 

[linebaugh]

very disturbing

 

[catalog]

 

[catalog]

 

[william_kent]

I will face my fear. I will permit it to pass over me and through me. And when it has gone past, I will turn the inner eye to see its path.

 

[william_kent]

this one genuinely surprised me - feeding H.P. Lovecraft quotes in to a GAN as a prompt "in the style of studio ghibli" and is that the name of a president I can see? has the GAN broached the wall separating us from true reality? is the name it has revealed " the "formless, infinite, unchanging and unchangeable void"?



To be bitter is to attribute intent and personality to the formless, infinite, unchanging and unchangeable void

 

[Clinamenic]

@catalog I think the target image parameter brings this to a whole new level.

We are one step closer to utilizing these GANs as narrative constructors. Depends on how closely the run can approximate the target image.

 

[william_kent]

@catalog I think the target image parameter brings this to a whole new level.

We are one step closer to utilizing these GANs as narrative constructors. Depends on how closely the run can approximate the target image.

it gets better - you can have arrays of target images

also, I found some code on GitHub that adds the caret ^ as a modifier for "story mode" - feed the GAN one phrase after another

jack and jill ^ went up the hill etc,

 

[Clinamenic]

also, I found some code on GitHub that adds the caret ^ as a modifier for "story mode" - feed the GAN one phrase after another

jack and jill ^ went up the hill etc,

Is it in this dummy-friendly notebook interface format? Or is it more of a manual thing?

 

[woops]

View attachment 8885

In his house at R'lyeh dead Cthulhu waits dreaming

View attachment 8890

In his house at R'lyeh dead Cthulhu waits dreaming

Best yet

 

[Clinamenic]

Ive been trying VQGAN+CLIP out on textual image inputs, a sort of procedural lettrism. This is the input image:





and here are a few of the results

with text input:
particle | waveform | electron microscopy
(seed 1 interval 2 iterations 100)

with text input:
floorplan | grid
(seed 1 interval 2 iterations 100)

with text input:
banknote | engraving | money
(seed 1 interval 5 iterations 250)
target image:

with text input:
city | streets | buildings
(seed 1 interval 5 iterations 250)
target image:

 

[linebaugh]

the last two are very good

 

[Clinamenic]

Yeah I think having the target image makes a big difference.

 

[catalog]

Good shit

 

[william_kent]

May the gods pity the man who in his callousness can remain sane to the hideous end!

 

[mixed_biscuits]

The most beautiful woman in the world. (https://hotpot.ai/art-maker)

 

[catalog]

used this initial image

with the text prompt

"tree man looking down | ent | HDR | hyperreal | photoreal

and got

 

[WashYourHands]

 

[william_kent]

Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn | in the style of Max Ernst sculpture

 

[catalog]

the magic mushrooms dance | in the style of james mcneill whistler

from this initial image