[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]

/lit/ - Literature

Search:


View post   

>> No.22342002 [View]
File: 1.92 MB, 406x247, cow-horse-in-a-field.gif [View same] [iqdb] [saucenao] [google]
22342002

>>22341820
Stable Diffusion.
See https://rentry.org/sd-mashup for full setup instructions.

>> No.22089740 [View]
File: 1.92 MB, 406x247, cow-horse-in-a-field.gif [View same] [iqdb] [saucenao] [google]
22089740

>>22089662
nickering

>> No.22069003 [View]
File: 1.92 MB, 406x247, cow-horse-in-a-field.gif [View same] [iqdb] [saucenao] [google]
22069003

>>22067982
This. Use AI-generated art as a beginning, not an end.
>>22068078
Stable Diffusion was trained by associating images with key words/phrases, and then devolving those images into noise.
It generates images by taking key words/phrases, generating a noise image, and then repeatedly denoising it until an image emerges.
Here's an example of "European style castle in Japan":
https://en.wikipedia.org/wiki/File:X-Y_plot_of_algorithmically-generated_AI_art_of_European-style_castle_in_Japan_demonstrating_DDIM_diffusion_steps.png
The AI knows what European style castles look like, and what Japan looks like, so by applying both prompts simultaneously, produces a blend.
Here's one where it alternately applies "cow" and "horse" in alternate generations, producing a hybrid cow/horse:
https://user-images.githubusercontent.com/39339941/197556926-49ceb72b-daf3-4208-86f3-c2e7e9cd775a.gif
Hopefully this is enough to get started.
>>22067137
"Artists" who seethe about AI are the ones who fear being replaced by it, and were never very good to begin with.

Navigation
View posts[+24][+48][+96]