Kharmakazy posted:

There's a lot of the names there

Tree Reformat posted:

feels a bit sacrilegious that this thread now probably constitutes the most effort ever put forth in making an actual zybourne clock deliverable, even jokingly

Ruffian Price posted:

A magic box that will happily invent nonexistent language features whole cloth, because the prime directive is to be helpful and satisfy your query. Verify against a search engine and use the feedback feature if you got a perfect-looking answer that doesn't actually work.

Doctor Zero posted:

That result absolutely stunned me. The rest are not quite so perfect, but awesome nonetheless less.

pixaal posted:

I don't get it how is their shining example not mona lisa or that it steals the style of Van Gogh. Like you could probably convince a jury that something there is wrong.

Ornamental Dingbat posted:

They made a sequel.

https://twitter.com/keithscho/status/1615387611916558336?s=20

PoorWeather posted:

Anyway, I don't think there's anything wrong with using open source stuff, or just loving around with the technology generally. The people I'm worried about are the ones who seem eager to stan the corporations behind this stuff and their products as if they're synonymous with it. I saw some people on a discord the other day basically worshipping Midjourney as a newborn god and it scared the hell out of me.

RPATDO_LAMD posted:

the gpt model is pre-trained, it is not learning from your inputs in real time or w/e. after MS's Tay companies aren't dumb enough to do that again
that "knowledge" only sticks around as long as the individual conversation thread it's in does.

IShallRiseAgain posted:

That looks like the old rich guy in the wheelchair in the Hannibal movie Red Dragon.

StarkRavingMad posted:

I would appreciate any tips on trying to generate pictures that look like a movie still frame or a photograph and less like a drawing. I'm trying some of the obvious things like "photorealistic" and "movie still frame" but it still kind of looks like an artwork. Or does it really just come down to the model you use?

I haven't messed around with Stable Diffusion in awhile and man, the advances are pretty impressive, especially screwing around with some of the custom models out there.

KwegiboHB posted:

Birthday request: everyone post monster trucks. Thank you!

pixaal posted:

Talking about building a large scale game with current tech is silly you have no room for anything but the AI and will require very powerful GPUs the cost of production and how niche a product this will be with most people going "yeah but it looks like poo poo and I need a 4090", this is 5-10 years out at a min, all of these features are things that have to be solved before it's viable and they can be.

I think currently you have to check state which means it's going to take multiple requests it will be slow like a DOS RPG, please enjoy this midi file while this loads for 15 minutes.

IShallRiseAgain posted:

I uh made a body horror LORA, and um well it certainly was effective. Even I was grossed out a bit by it. You can get it here https://civitai.com/models/18399/body-horror

Roman posted:

https://twitter.com/RomanIsntOnline/status/1641749709525069824

Doctor Zero posted:

Wow, that's a great idea!


Wait... I use AI art because I CAN'T draw!

KwegiboHB posted:

Midjourney Img2Txt



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

deep dish peat moss posted:

TIP has been working with a custom text2animation model ITT, reading their posts may give a better understanding of the kind of effort you're looking at - here's the most recent:

Pure speculation from me but: It'll probably be another year or two (assuming no major roadblocks in AI animation show up) before AI will be able to realistically do a passable version of the kind of animation you're looking for.


A "more realistic" way to do it (right now) would be hand-drawing an outline for each frame and then using SD's capability to use that to guide the shape of the generation to fill it in (I forget the name of that tool, not a SD person myself). That's realistically "small team" work but an enormous amount of work hours for the individuals (sketching and generating each frame individually, stitching them together into an animation)



edit: But also as an incredibly sincere suggestion for small-team low-effort animation, consider how Squigglevision works

The only thing you need for something to 'feel' animated is kinetic movement, even if there's no actual animation. You can easily turn a series of still images into something that feels animated, without actual animation. This type of animation (not Squigglevision specifically, but "mostly still images with kinetic movement") worked well for tons of Adult Swim shows, video games (think Darkest Dungeon), digital versions of comic books, etc. In game development it's generally called "Juice" and you can find plenty of videos/articles about how to add "juice" to your game that could give you some ideas which could be translated to animation. So don't let the accessibility of Animation turn you away from following your creative dreams You could realistically do something with this concept using AI and a little bit of Human elbow grease right now.

deep dish peat moss posted:

Also I haven't drawn much lately because I'm still waiting for a change in medication to even out but here's an example of the kind of thing I've been doing with AI recently:

1) AI-generated concept image:

(This was MJv4)

Crapple! posted:

quote:

...

Bucnasti posted:

It doesn't understand anything, it associated non-english words with images based on it's training data. Which is why when you feed it prompts in German, it spits out images that look like they're from Germany, because the training data images that were labeled in German were most likely taken in Germany.
English results in more varied output because more of the training data came from English sources.

AARD VARKMAN posted:

im game for hero jobs as a theme

feedmyleg posted:

I've used a combo of getimg.ai+dreambooth and MidJourney in two video projects over the past week and have been totally blown away with how much time they're able to save me and how much more creativity they're able to unlock. Especially when I can use Photoshop to make a mock-up of what I want, then use getimg.ai's text-to-image to hone it into the style I want in 5 seconds. Making a music video where everything was both custom and convincingly watercolor in a very short amount of time was an absolute breeze.

Roman posted:

Making some previs/promo shots for a character. It's rare I get a bunch of stuff this good in one session. Based on Fiona Dourif, with bonus dollar store Elijah Wood at the end.

Brawnfire posted:

My prompt here ended up as "cupped hands full of minestrone no bowl" and it only worked once out of dozens of iterations

Brawnfire posted:

If you want a challenge, make a cupped pair of hands holding soup without a bowl

My prompt here ended up as "cupped hands full of minestrone no bowl" and it only worked once out of dozens of iterations

Fuschia tude posted:

Many hands make light work

Bucnasti posted:

Trying to make some NPC images for Dungeons and Dragons and for some reason, Bing is convinced that dwarf women have big pointy ears, and the more I try to convince it otherwise the bigger it makes them.

Brawnfire posted:

pixaal posted:

GPT is fine at like an adult swim style for about 3-5 minutes tops then it goes off the rails. I really do think current AI could be curated (this part is important!!) into a viable 2AM sketch show with 5 minute long sketches without more than an improv line read edit. Oh you want the AI to read it too, and decide the emotional tone? Oh boy now you are just being wacky. We might be able to get 15-30 second long consistent sketches. It can do the couch gag equivalent to something else.

Gromit posted:

Agreed. I'm not into model trains at all, but I liked some other people's work that had little model houses and plastic people, so wanted to see what AI would do with a whole train set. I'd love to see dioramas of famous historical events, but couldn't think of any in particular that would be fun to see.

Soulhunter posted:

be the change you want to see in the world and make a website with gpt

The mini not-quite-daily themes here have kind of covered the Photoshop Phridays concept, but as for a dedicated site, I'm not aware of one outside of the "Top" queue with hidden voting algos for Midjourney or posting on the usual social media and counting the likes. There's probably a bunch of scattered webui stable diffusion sites with that concept built-in.

I'd be down for a weekly / themed thread on the forums here if there was enough interest.

AARD VARKMAN posted:

Mola Yam posted:

BoldFace posted:

Marvel made the intro to their new television show using AI.

https://twitter.com/PrimeTimeFilms_/status/1671441123259105280

Roman posted:

It's also likely that it was made a while ago with older tech. Progress is measured in months with this stuff.

Crapple! posted:

Thanks I'll have to give img2img a shot sometime. I definitely get the panning for gold analogy. The one time I got something like that picture of yours was when I generated this from a chaos 100 prompt



And then I used the zoom function

Mercury_Storm posted:

Got some high quality bathroom art now

unzin posted:

2 years ago I was playing with CLIP+VQGAN... can't find one working anymore, but I liked the abstractness of it.

https://i.imgur.com/uDWS2Xy.mp4

Mercury_Storm posted:

Fraudo Sam the Hobbit on the witness stand in front of a jury playing League of Legends on a gamer laptop in a court room sketch

Debunk This! posted:

I'm sure this is pretty boring to most of the thread regulars but I'm having fun trying to create specific fantasy characters and scenes. Anyways heres some Sabriel.