Uriah Heep posted:

[...]

Cabbages and Kings posted:

[...] yeah I couldn't look at trumbo anymore, I had to go make this to calm down after

BoldFace posted:

https://twitter.com/weirddalle/status/1603023590303338498

SniperWoreConverse posted:

[...]

Megazver posted:

I don't find Clip Interrogator terribly useful, usually, unfortunately.

These images are very cool. You might find 'Bernie Wrightson' useful to add into the mix for that kind of art. Junji Ito I think might be a little too recognizable. Like, these are obviously, blatantly Junji Ito pastiche to the point where I'd personally be uncomfortable using them in a commercial product:



Yeah, it's not enough for the AIs. I've been experimenting with different phrasings, but it's not easy.

KillHour posted:

Sourdough Sam posted:

Posting will be the last true art, and it's not looking good for us either.

Tiny Myers posted:

I guess the thing that differentiates it for me is that these AIs are so increasingly sophisticated. Like, I think something like ChatGPT would fool a lot of people (not extremely-online people, just random people off the street who aren't very tech-savvy, your grandma or something) into thinking it's a real person if not for constantly reassuring you it isn't. I had a conversation on https://beta.character.ai/ that felt remarkably convincing. Left-clicking to pop bullets at a floppy facsimile of a nazi in a game where you logically know their AI doesn't go beyond game mechanics (not to cheapen the work of AI designers for games, I know the programming is much more sophisticated than this, I just mean the complexity is obviously much different from something like GPT) feels a lot different from the robot that can write me a beautiful poem about itself just because I asked. It feels closer to wearing a VR helmet and shooting at people with realistic faces who cry and beg for their lives. Yeah, you know they're digital, but...

I'm not foolish enough to think that the AI has actual sentience of any kind or feels pain or other emotions, but I can't empathize at all with doing things like that, and it kind of disturbs me that so many people do.

But like, I'm not trying to navel gaze too much, it's a genuine question, and I think you have good points. I've wondered that, if maybe it's good in the long run to have things like this. Like maybe if someone simulates some horrible thing, they'll be satisfied and/or decide "actually this sucks and I have no desire to do it in real life". But on the other hand, armies use VR training for its soldiers.

To be very clear, I'm not trying to doomsay oohhh ai is bad and we should stop making it. I love AI advances and all the creative things people are doing with it, even as a career artist (a lot of my peers are really upset about AI art that works off of datasets scraped from the internet without attribution). It's more "I hate that every time I read about AI I discover some new and creatively upsetting thing other human beings are doing with it"

cinnamon rollout posted:

Maybe I am just bad at AI, but I've found it very difficult to change specific things to other very specific things. For example changing the color of a short from like grey to blue. Stable diffusion just can't seem to get it done, and if I want a very specific color of blue it's impossible.
I guess I would need to do something like gather a handful of pictures of the specific color shirt I want, and train it on those?
I've had fun making random stuff, but getting specific has been pretty frustrating.

BARONS CYBER SKULL posted:

What’s more ethical

threatening someone to get them to do what you want

Or

Installing bolts into someone’s brain to restrict what they can think

BrainDance posted:

I wrote a very long, detailed, step by step guide to finetuning the larger GPT-Neo models including text formatting. There was a lot of stuff that just wasn't well documented that I tried to document. I host it on my personal website though so I don't want to link it (its got my name in it) but, would people be interested in that and mind if I make a post or two out of it? Like I said it's long so I don't wanna just go and do it.

I trained GPT-Neo 1.3B on the Laozi, Liezi, and Zhuangzi (the 3 big Daoist philosophers) and my results are actually better than I thought

KPC_Mammon posted:

My girlfriend is playing a goblin in our Pathfinder game and we were having a hell of a time using stable diffusion to make anything resembling a goblin. I wish people included prompts so I could make a less sexually objectified version that actually has a face. I should ask the TG thread for help making fantasy stuff.

Sexual Lorax posted:

[...]
one of the party tricks id like to get better at is changing the style of a photo to oil painting or dagguerotype or whatever and i know that img2img in automatic1111 is the core of that but this tech moves so quickly that im having one devil of a time finding documentation or tutorials so if someone has the time and inclination to point me in a direction id be greatly appreciative
[...]

busalover posted:

New tool, not sure how much benefit it brings over the existing ones

Composer: Creative and Controllable Image Synthesis with Composable Conditions
https://damo-vilab.github.io/composer-page/
https://github.com/damo-vilab/composer

Nigmaetcetera posted:

Can you make it cry, I ask for non-sadistic reasons?

Edit: drat it I meant to hit back, not post! I don’t want to torture an AI, just bully it a little, I swear.

Junk posted:

built myself a new pc and decided to give stable diffusion another shot. I didn't realize that it was so dependent on nvidia, which sucks because i bought a radeon 7900xt. turns out there's a fork of stable diffusion webui that uses direct ml, only downside is it eventually will tell me i've run out of video memory if I create enough images in a row. dunno if anyone has any advice to give on this that doesn't involve giving nvidia money.

Gashroom posted:

Blobfish goes shopping

Dead Sex-Parrot posted:

prompt:

code:

thank you :)

Small Strange Bird posted:

A big old 'gently caress you' to whoever trains Bing/Dall-E: I was trying to get a picture of a woman of, y'know, average build rather than the rail-thin models AI image generators tend to default to when you ask for "a woman". The problem with Bing is that most of the descriptors you'd use to do that get eggdogged, but I eventually managed to get "slightly overweight" to work in a prompt.



This is Bing's idea of a "slightly overweight" woman - ie, perfectly normal and healthy.