Stable Diffusion, NovelAI, Machine Learning Art - AI art generation discussion and image dump

[Lord Xenu]

Could you use and download this on a Mac?

There is a script here for automated installation on newer Mac's. I haven't used it, I just saw it mentioned. I assume if you have an intel CPU it will be restricted to CPU only, and it looks like support Apple silicon (Mx) is limited. If you look at the comments at the end of this post, at least one person was able to get it working on an Intel Mac.

 

[The Last Stand]

There is a script here for automated installation on newer Mac's. I haven't used it, I just saw it mentioned. I assume if you have an intel CPU it will be restricted to CPU only, and it looks like support Apple silicon (Mx) is limited. If you look at the comments at the end of this post, at least one person was able to get it working on an Intel Mac.

I have an M1 chip MacBook Air. 2020. It should work.

 

[Glowie]

Banned!

View attachment 3753294

Artist meltdown has begun, sit back and watch it spread

 

[Bananadana]

this isn't what I meant by "globohomo art"
View attachment 3753462

The powerword for globohomo art is (Corporate Memphis), ideally with "illustration"
These are all text generated from Huggingface with the keywords:
masterpiece, best quality, (Corporate Memphis), illustration, Man holding a gun
With 42 sample steps.


Somehow still better than the actual corpo stuff, but even the AI apparently wont fuck up its art that much.

 

[Puff]

I kinda want to train a model on my photos, but I have a feeling that this would require about a work year of tagging.

 

[IamnottheNSA]

The powerword for globohomo art is (Corporate Memphis), ideally with "illustration"
These are all text generated from Huggingface with the keywords:
masterpiece, best quality, (Corporate Memphis), illustration, Man holding a gun
With 42 sample steps.
View attachment 3753514View attachment 3753515
View attachment 3753516View attachment 3753518
Somehow still better than the actual corpo stuff, but even the AI apparently wont fuck up its art that much.

Unlike the abomination that is globohomo art these actually look pretty neat. They remind me of Microsoft Word 97 clipart.
I should try using (((allegria))) as the powerword.

 

[The Last Stand]

The powerword for globohomo art is (Corporate Memphis), ideally with "illustration"
These are all text generated from Huggingface with the keywords:
masterpiece, best quality, (Corporate Memphis), illustration, Man holding a gun
With 42 sample steps.
View attachment 3753514View attachment 3753515
View attachment 3753516View attachment 3753518
Somehow still better than the actual corpo stuff, but even the AI apparently wont fuck up its art that much.

Can you do one with Bette Davis, please?

 

[Lord Xenu]

I kinda want to train a model on my photos, but I have a feeling that this would require about a work year of tagging.

I want a South Park specific model. I tried using NAI a few times and only got nonsense in return. Its kind of surprising to see South Park studios hasn't tried to turn that into a product already, considering they had some AI enhanced youtube show a while back.

 

[Scavenger]

We're FUCKED lmao.
Being an artist is gonna be so hard now
Espeically on the internet.

Maybe this is a weird wake up call to diffrenciate your artstyle or mix it up and also;
Don't be a dick to an artist you know going "HAHA LOL AI IS BETTER!!11".
cut ya local artist some slack. They need it

 

[MarvinTheParanoidAndroid]

On average, life just gets better. I'm fucking old. When I was born, there were no computers. Things sucked in many ways. A close relative of mine died of cancer when I was a kid. They didn't even know exactly what kind of cancer, or where it started. Diagnostics just weren't that advanced. My father later died terribly of a disease he most likely would've been cured of if he got it now. Why? Technology. As simple as. I'm pretty sure a few of us will die of shit people 30 years later will marvel at when they hear people actually used to die of it and that simply was the way it was.

It's easy to look back with rose-tinted glasses and feel uncomfortable when faced with change, as often things seem to be just fine now, and I'm prone to do so myself often enough, but progress so far has been a net-positive. Yes, I'm almost certain AI will be abused for terrible shit. I also believe it'll genuinely improve, probably even save, many lives. Things seldom are binary like that.

Do you suppose that stable diffusion could somehow be used to treat or cure cancer? That's a use case I could get behind, although I have no idea how you could use it for that. I don't think it was meant for that or it could do that so I have my doubts.

 

[Stalphos Johnson]

Do you suppose that stable diffusion could somehow be used to treat or cure cancer? That's a use case I could get behind, although I have no idea how you could use it for that. I don't think it was meant for that or it could do that so I have my doubts.

I doubt stable diffusion itself could, but I know there are a few image recognition AIs floating around that can identify cancer cells from normal cells on histology slides with similar accuracy to most pathologists. It's going to be curious to see how AI affects healthcare. I have a feeling it's going to be harder to get a pathology residency or a radiology residency unless you're planning to do interventional radiology.

 

[Eddie Riggs]

Politicians in 3 artist's styles

 

[MarvinTheParanoidAndroid]

I doubt stable diffusion itself could, but I know there are a few image recognition AIs floating around that can identify cancer cells from normal cells on histology slides with similar accuracy to most pathologists. It's going to be curious to see how AI affects healthcare. I have a feeling it's going to be harder to get a pathology residency or a radiology residency unless you're planning to do interventional radiology.

Medicine is the one field I can think of that would benefit significantly from AI or automation without putting people out of work, simply because it would make it more empirical and help weed out corruption and malpractice.

 

[themasterlurker]

V1.5 is now out

https://huggingface.co/runwayml/stable-diffusion-v1-5/tree/main

 

[Mr.Miyagi]

Some interesting drama apparently going on with 1.5, as RunwayML put it onto HuggingFace, only for StabilityAI to request a takedown, claiming it was a leak of their IP. Runway's CEO made a statement in that discussion about SD's open-source license, and it looks like Stability's legal team went and reverted the takedown request. Seems like there's been plenty of drama going on in Discord and Reddit spaces, but I don't know all the details, maybe somebody else can put all the pieces together.

Anyways, it looks like version 1.5 was trained on 595k steps from v1.2, with 1.4 being just 225k steps from 1.2. It'll be interesting to see if this causes any significant improvement or change in generation quality. If you want to grab 1.5 directly, keep in mind you'll need to visit the main 1.5 page, have a HuggingFace account, and agree to the terms and conditions and all that jazz. Looks like there's a torrent for it here.

 

[themasterlurker]

Some interesting drama apparently going on with 1.5, as RunwayML put it onto HuggingFace, only for StabilityAI to request a takedown, claiming it was a leak of their IP. Runway's CEO made a statement in that discussion about SD's open-source license, and it looks like Stability's legal team went and reverted the takedown request. Seems like there's been plenty of drama going on in Discord and Reddit spaces, but I don't know all the details, maybe somebody else can put all the pieces together.

Anyways, it looks like version 1.5 was trained on 595k steps from v1.2, with 1.4 being just 225k steps from 1.2. It'll be interesting to see if this causes any significant improvement or change in generation quality. If you want to grab 1.5 directly, keep in mind you'll need to visit the main 1.5 page, have a HuggingFace account, and agree to the terms and conditions and all that jazz. Looks like there's a torrent for it here.

People are reporting that smaller details are nicer, and hires fix is not as necessary anymore, slightly more comprehensive vocabulary, etc

 

[Gaunt King Ithan Rilph]

Been fucking around with my pfp and inpaint/img2img.

 

[Baraadmirer]

Some interesting drama apparently going on with 1.5, as RunwayML put it onto HuggingFace, only for StabilityAI to request a takedown, claiming it was a leak of their IP. Runway's CEO made a statement in that discussion about SD's open-source license, and it looks like Stability's legal team went and reverted the takedown request. Seems like there's been plenty of drama going on in Discord and Reddit spaces, but I don't know all the details, maybe somebody else can put all the pieces together.

Anyways, it looks like version 1.5 was trained on 595k steps from v1.2, with 1.4 being just 225k steps from 1.2. It'll be interesting to see if this causes any significant improvement or change in generation quality. If you want to grab 1.5 directly, keep in mind you'll need to visit the main 1.5 page, have a HuggingFace account, and agree to the terms and conditions and all that jazz. Looks like there's a torrent for it here.

Can I update it with a git pull or do I have to download it directly?

 

[Chicken Gendies]

It's really easy to waste alot of time on generating images, I didn't get much work done yesterday because of it. I really like making desktop wallpapers, but it seems to have trouble upscaling darker images without altering the original too much or putting black smears over it.


Here's one it struggled with:

Tried using 4x_foolhardy_remacri and lollypop but they both produced similiar results.

 

[A Hot Potato]

Been messing around with the red haired (now orange haired) girl.