Live data from Hacker News

How a Stable Diffusion prompt changes its output for the style of 1500 artists

gorgeous.adityashankar.xyz

81–90 of 205 posts

Re: How a Stable Diffusion prompt changes its output for the style of 1500 artists

#81

Earlier quoted context omitted.

Flyers are an extremely low bar that AI sails over.

show an example?

I used to draw the flyers when my band played shows. People still went to the shows. If had iterated a few times with one of these models to draw the thing that I wanted to draw rather than deluding myself, and spent 5 minutes polishing it in photoshop, the product would have been orders of magnitude better imo.

Re: How a Stable Diffusion prompt changes its output for the style of 1500 artists

#83
post #69

It would be so much better to just have all of these 1500 pictures next to each other in a list with the artist's name under it.

Or at the very least with the list of artists alphabetically sorted!

They're sorted by last name, it took me a second too! I just wish I knew more of them.

Re: How a Stable Diffusion prompt changes its output for the style of 1500 artists

#84

What frustrates me about Stable Diffusion is there doesn't seem to be any documentation as to what artists or vocabulary it understands. Generally people say "look at existing prompts or use various prompt generators" but that doesn't really solve the problem. I don't want to just look at what other people have randomly discovered; I want to know what the program really knows.

"It" doesn't understand anything.

It's just a very advanced madlibs engine based on a database of a billion of already known images.

Re: How a Stable Diffusion prompt changes its output for the style of 1500 artists

#85
post #20

Earlier quoted context omitted.

A good measure for whether you're more of a celebrity or an artist is how much of your face a google-image trained AI thinks belongs in your work.

Well what if the artist has a popular self portrait?

Frida Kahlo comes to mind.

Re: How a Stable Diffusion prompt changes its output for the style of 1500 artists

#86

Truly impressive, though it gives me some great fear for all the people whose career is art. This will likely take the pareto to a new height instead of 20/80 maybe even 1/99

After checking some 19th century artists I see it failed hard. All of the responses look the same, there isn't enough data in the training set to differentiate actual styles beyond "vaguely realistic".

Re: How a Stable Diffusion prompt changes its output for the style of 1500 artists

#87
post #67

What's kind of crazy is how the images tend to have similarities in small features that become very apparent when flipping back and forth between images, but which are not obvious per se. For example, I flipped back and forth between Beatrix Potter and Paulus Potter. A rounded white bonnet in one picture becomes a couple of blossoms in the other. The roof of a house becomes some shadowy wall with plants in the other.…

So interesting, and really seems somehow related to how we dream, hallucinate.. or even experience reality?

Re: How a Stable Diffusion prompt changes its output for the style of 1500 artists

#88

Very cool! I recently made a game kind of like AI "pictionary" where the user has to guess the "artist", subject, and description of a piece of art generated by stable diffusion: https://wcedmisten.fyi/project/paintingGuesser/ I tried to make something more general, but stable diffusion is fairly inconsistent in how well the output matches the semantics of the input.

This could be a good candidate for a wordle-style scoring system

Re: How a Stable Diffusion prompt changes its output for the style of 1500 artists

#89

Hate to say it but when i see stuff like this it only reminds me of what we could have achieved if this ingenuity had been applied in another domain. Can't help feeling that this accidentally harms creative types and risks swamping us with visual junk. The technical achievment is astounding but no-one would seriously claim that crafting an image via a short prompt is creative except in the most cursory way. I'm proba…

I don't know why everyone assumes that ML researchers have some big map of the future where they can make decisions like "yeah, let's choose this branch over here, where AI gets good at generating art first, rather than that other one where it cures cancer a decade earlier." The breakthroughs come where they come, and no one knows where some model architecture will have an application in the future.
Post reply on HN