Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Post
  • Reply
Lord of the rivets
Sep 10, 2022

Swagman posted:

I dont know if this has been posted yet, but it has a nice list of artists and style examples.

https://supagruen.github.io/StableDiffusion-CheatSheet/


This is a great resource :cool:

style of Robert Delaunay, world war 2 tank battle


style of Lisa Frank, world war 2 tank battle

Adbot
ADBOT LOVES YOU

Swagman
Jun 10, 2003

Yes...all was once again peaceful in River City.






















Sedgr
Sep 16, 2007

Neat!

Soulhunter
Dec 2, 2005






Roman
Aug 8, 2002

edit: nm for now

Roman fucked around with this message at 10:19 on Sep 19, 2023

tango alpha delta
Sep 9, 2011

Ask me about my wealthy lifestyle and passive income! I love bragging about my wealth to my lessers! My opinions are more valid because I have more money than you! Stealing the fruits of the labor of the working class is okay, so long as you don't do it using crypto. More money = better than!
So, uh, this technology is actually extremely impressive, but, uh, how are the original artists ,that all this training data is based on, being compensated? Is that even possible?

KakerMix
Apr 8, 2004

8.2 M.P.G.
:byetankie:

tango alpha delta posted:

So, uh, this technology is actually extremely impressive, but, uh, how are the original artists ,that all this training data is based on, being compensated? Is that even possible?

You're a court case dismissed and a year too late for that debate

deep dish peat moss
Jul 27, 2006

tango alpha delta posted:

So, uh, this technology is actually extremely impressive, but, uh, how are the original artists ,that all this training data is based on, being compensated? Is that even possible?

The same way that Disney gets compensated when you draw a picture of Mickey Mouse

e: More seriously it's pretty much the same concept as traditional art. Using someone else's work or style as an influence on your own doesn't give them any claim to your work. But if you're trying to pass it off as being by them, profiting off of content that they hold the copyright to, or if they can prove that your work has negatively impacted them, then they have a case against you. It's arguable whether that makes AI art ethical but there's no precedent in the art world that artists get paid when others copy their style. If you're pumping out AI-generated images of Mickey Mouse, Disney can come after you for your use of the Mickey Mouse character, but they have no claim against you if you're generating wholly original content in the style of a disney animation.

This is all of course subject to change as AI evolves and things go to court but as far as how it works right now, that's about it :shrug:

deep dish peat moss fucked around with this message at 08:53 on Sep 18, 2023

Roman
Aug 8, 2002

Because of all the murky copyright issues, right now I'm making AI image/video and related stuff for free. I look at it like making a mod for a game. Exceptions like Black Mesa aside, I don't have retail product plans in mind. I might do the Patreon thing for people who would like to support me making stuff, but that's it for now.

(If I got a LOT of money, from investors or whatever, honestly I wouldn't even have to be using AI. I'd just film a normal film.)

Roman fucked around with this message at 12:45 on Sep 18, 2023

leper khan
Dec 28, 2010
Honest to god thinks Half Life 2 is a bad game. But at least he likes Monster Hunter.

deep dish peat moss posted:

The same way that Disney gets compensated when you draw a picture of Mickey Mouse

e: More seriously it's pretty much the same concept as traditional art. Using someone else's work or style as an influence on your own doesn't give them any claim to your work. But if you're trying to pass it off as being by them, profiting off of content that they hold the copyright to, or if they can prove that your work has negatively impacted them, then they have a case against you. It's arguable whether that makes AI art ethical but there's no precedent in the art world that artists get paid when others copy their style. If you're pumping out AI-generated images of Mickey Mouse, Disney can come after you for your use of the Mickey Mouse character, but they have no claim against you if you're generating wholly original content in the style of a disney animation.

This is all of course subject to change as AI evolves and things go to court but as far as how it works right now, that's about it :shrug:

USPTO is working out their position right now. They could very well determine that ai generated works are derivative works of every work in the training set. We just don't know.

hydroceramics
Jan 8, 2014
Some more spooky photos





















Vlaphor
Dec 18, 2005

Lipstick Apathy
Having a bit of fun making spooky places.







I tried to make a house by the cemetery and it just put the cemetery at the house

Count Roland
Oct 6, 2013


One of my favorite images I've seen in this thread. Very unique. Looks like a poster for a Tim Burton movie, but an actually good one.

pixaal
Jan 8, 2004

All ice cream is now for all beings, no matter how many legs.


making some weird rear end poo poo with Bing

kitchen aid mixer gone wrong, artist Lisa Frank, theme potato, 4k, live action

kitchen aid mixer gone wrong, artist Guy Fieri, theme of flavor

kitchen aid mixer gone wrong, artist Willy Wonka, theme vampires (Vampires do not eat limbs! bad bing!)

kitchen aid mixer gone wrong, artist Old Greg, theme Baileys

kitchen aid mixer gone wrong, artist some kitten, theme citrus

pixaal fucked around with this message at 17:38 on Sep 19, 2023

LifeSunDeath
Jan 4, 2007

still gay rights and smoke weed every day
https://clips.twitch.tv/StrongTubularDillHotPokket-RAcYHyP01MAmT-3O

Doctor Zero
Sep 21, 2002

Would you like a jelly baby?
It's been in my pocket through 4 regenerations,
but it's still good.

Artists being compensated for using their work in training data.



(Dunno how that works but I thought it was a neat image)

Spoopy





pixaal
Jan 8, 2004

All ice cream is now for all beings, no matter how many legs.


I like the concepts for brain dance, brain freeze

AARD VARKMAN
May 17, 1993

Swagman
Jun 10, 2003

Yes...all was once again peaceful in River City.


























Roman
Aug 8, 2002

edit:nm

Roman fucked around with this message at 01:45 on Sep 26, 2023

XYZAB
Jun 29, 2003

HNNNNNGG!!
Pack it up guys, we’ve finally achieved what was once considered impossible and future generations will never forgive us for it.

https://www.youtube.com/watch?v=Fop_FBc_jPA

Doctor Zero
Sep 21, 2002

Would you like a jelly baby?
It's been in my pocket through 4 regenerations,
but it's still good.

:gaz:

Now we finally know the true cause of WWII. Somehow that makes it back in time, Hitler hears it and :boom:

Hadlock
Nov 9, 2004

https://openai.com/dall-e-3

DALL-E 3 announced/released(?)

compare 2 vs 3:



quote:

DALL·E 3 understands significantly more nuance and detail than our previous systems, allowing you to easily translate your ideas into exceptionally accurate images.

DALL·E 3 is built natively on ChatGPT, which lets you use ChatGPT as a brainstorming partner and refiner of your prompts. Just ask ChatGPT what you want to see in anything from a simple sentence to a detailed paragraph.

DALL·E 3 in ChatGPT

When prompted with an idea, ChatGPT will automatically generate tailored, detailed prompts for DALL·E 3 that bring your idea to life. If you like a particular image, but it’s not quite right, you can ask ChatGPT to make tweaks with just a few words.

DALL·E 3 will be available to ChatGPT Plus and Enterprise customers in early October. As with DALL·E 2, the images you create with DALL·E 3 are yours to use and you don't need our permission to reprint, sell or merchandise them.

and then this near the bottom

quote:

DALL·E 3 is designed to decline requests that ask for an image in the style of a living artist. Creators can now also opt their images out from training of our future image generation models.

notes from the peanut gallery:

quote:

https://news.ycombinator.com/item?id=37587836

This is the first time they have not posted any sort of a paper when releasing a new model. Even the sort that accompanied GPT-4 announcement.

hydroceramics
Jan 8, 2014

Hadlock posted:

https://openai.com/dall-e-3

DALL-E 3 announced/released(?)

compare 2 vs 3:



and then this near the bottom

notes from the peanut gallery:
I wonder what their pricing model will be. Dalle2s was not realistic or sustainable.

Tunicate
May 15, 2012

hydroceramics posted:

I wonder what their pricing model will be. Dalle2s was not realistic or sustainable.

attached to your chatgpt subscription apparnetly

smoobles
Sep 4, 2014

Tunicate posted:

attached to your chatgpt subscription apparnetly

that's... not bad

hydroceramics
Jan 8, 2014

Tunicate posted:

attached to your chatgpt subscription apparnetly

I had to dig for it (great ux!) but that seems to be 20usd/mo. Getting gpt4 and dalle3 for that seems like a bargain, especially if they're integrating them. Dalle3 has a lot of ground to cover to be anywhere as good as the competition at this point, though.

Hadlock
Nov 9, 2004

OpenAI's business model seems to be

1) be miles ahead of everyone else
2) offer professional services to tune the model to your specific appication/budget

smoobles
Sep 4, 2014

are some AI arts real?

KakerMix
Apr 8, 2004

8.2 M.P.G.
:byetankie:

Hadlock posted:

OpenAI's business model seems to be

1) be miles ahead of everyone else
2) offer professional services to tune the model to your specific appication/budget

They start milea ahead, but they are losing ground very quickly, if already the lead, in LLM performance. They had the big splash on Dall E II but quickly lost to both Midjourney and Stable Diffusion, to say nothing about Adobe.

We'll see how Dall-E 3 is, maybe there is some cool aspect about it.

Hadlock
Nov 9, 2004

Yeah based on what I am observing here midjourney appears to be the market leader right now at least in terms of "wow factor"

Integration with Chat-GPT is probably a "killer feature" though, everyone "gets" how to use chat-gpt. You can create some hekkin rad images with midjourney and sd-xl but many times you need a complex render chain with a bunch of positive and negative modifiers. Being able to say "can you do this, but make it more pink shaded?" then "oh I wanted the previous image, but only like 10% more pink. this is too pink" is something my mom can pick up and use

In other news, toyota released this today, applying diffusion models on 3d space tasks for robot arms. It's pretty high level but looks very promising

https://www.youtube.com/watch?v=w-CGSQAO5-Q

hydroceramics
Jan 8, 2014

KakerMix posted:

They start milea ahead, but they are losing ground very quickly, if already the lead, in LLM performance. They had the big splash on Dall E II but quickly lost to both Midjourney and Stable Diffusion, to say nothing about Adobe.

We'll see how Dall-E 3 is, maybe there is some cool aspect about it.

I'd put the fully free version of gpt on par with bard, which isn't saying much. Haven't tried the paid version.

smoobles
Sep 4, 2014

GPT 4 is scary good, and useful for a lot of menial tasks depending on your industry

KakerMix
Apr 8, 2004

8.2 M.P.G.
:byetankie:

Hadlock posted:

Yeah based on what I am observing here midjourney appears to be the market leader right now at least in terms of "wow factor"

Integration with Chat-GPT is probably a "killer feature" though, everyone "gets" how to use chat-gpt. You can create some hekkin rad images with midjourney and sd-xl but many times you need a complex render chain with a bunch of positive and negative modifiers. Being able to say "can you do this, but make it more pink shaded?" then "oh I wanted the previous image, but only like 10% more pink. this is too pink" is something my mom can pick up and use

In other news, toyota released this today, applying diffusion models on 3d space tasks for robot arms. It's pretty high level but looks very promising

https://www.youtube.com/watch?v=w-CGSQAO5-Q

Right it's the accessibility that OpenAI has with ChatGPT and Microsoft's deep investment that's really working for them. The dominance in the same way that smartphones were iPhones for the longest time to people that didn't know any better.

This whole space is still moving at light speed, it's not slowed down at all, maybe gotten a little stale news-wise. The sort of models I can run on my 4090 are wild as it is, to say nothing about renting GPUs from somewhere and running models on those.

Hadlock
Nov 9, 2004

Yeah in 10 years someone will design a CPU that can run sd-xl/gpt-3 that can run on 10w and fits in the thermal envelope of a golf ball and oh god I am describing the plot of west world aren't I. I swear I didn't mean to

feedmyleg
Dec 25, 2004

smoobles posted:

GPT 4 is scary good, and useful for a lot of menial tasks depending on your industry

As a creative writing assistant it's amazing. The idea of "Give me 10 ideas for a character backstory for an arrogant doctor who works at a mental institution and mistrusts police, make them 200 words each" gives a ton of inspiration at the snap of a finger. Taking bits and pieces from multiple options then feeding it back in and honing it works great, too.

KwegiboHB
Feb 2, 2004

nonconformist art brut
Negative prompt: amenable, compliant, docile, law-abiding, lawful, legal, legitimate, obedient, orderly, submissive, tractable
Steps: 32, Sampler: DPM++ 2M Karras, CFG scale: 11, Seed: 520244594, Size: 512x512, Model hash: 99fd5c4b6f, Model: seekArtMEGA_mega20

smoobles posted:

are some AI arts real?

Someone finally asking the big questions.

We could be here awhile in trying to answer them.


are some AI arts real?
Steps: 48, Sampler: DPM++ 2M Karras, CFG scale: 11, Seed: 1, Size: 512x512, Model hash: 88967f03f2, Model: juggernaut_final, Denoising strength: 0.48, RNG: CPU, Hires upscale: 2, Hires steps: 32, Hires upscaler: 4x-UltraSharp, Version: v1.5.2


are some AI arts real?
Steps: 48, Sampler: DPM++ 2M Karras, CFG scale: 11, Seed: 1, Size: 512x512, Model hash: 99fd5c4b6f, Model: seekArtMEGA_mega20, Denoising strength: 0.48, RNG: CPU, Hires upscale: 2, Hires steps: 32, Hires upscaler: 4x-UltraSharp, Version: v1.5.2


are some AI arts real?
Steps: 48, Sampler: DPM++ 2M Karras, CFG scale: 11, Seed: 4, Size: 512x512, Model hash: f65275abcf, Model: notreallyistic_v2, Denoising strength: 0.48, RNG: CPU, Hires upscale: 2, Hires steps: 32, Hires upscaler: 4x-UltraSharp, Version: v1.5.2



Maybe the better question is, How real do you want it to be?

https://www.poppy-project.org/en/

AARD VARKMAN
May 17, 1993
Holy poo poo, this is a whole different level of ability to combine concepts in one prompt.

https://twitter.com/sama/status/1704547625482203560
https://twitter.com/sama/status/1704547627583553961

Doctor Zero
Sep 21, 2002

Would you like a jelly baby?
It's been in my pocket through 4 regenerations,
but it's still good.

If it will give me a picture of a 1920s gangster holding an ugly coffee mug, I will hail it as lord and savior.


Fake edit: working text :swoon:

Adbot
ADBOT LOVES YOU

Hadlock
Nov 9, 2004

Doctor Zero posted:

Fake edit: working text :swoon:

Working text might be the thing that finally convinces the unbelievers

Probably why they said "no living artist" is they don't want someone to turn around and generate a bunch of Calvin and Hobbes comic strips and sell it as "the secret 4th edition of the anthology" or something

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply