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

[SomethingProblematic]

The second one was the icon as image, but description was "ugly asian man in a wig".
Using the same description and your avvy I get:
View attachment 3750666
But if I try "sassy anime nazi" it's:
View attachment 3750668


Why not combine the two?
Input:
View attachment 3750674
Result:
View attachment 3750680View attachment 3750681


View attachment 3750687


These are horrible.
View attachment 3750708View attachment 3750709


https://docs.novelai.net/

Do it with the non anime generator thing

That's super cute! What settings do you use? When I input it, it did nothing to it except make the hair a bit more detailed. Here are some dog requests:
View attachment 3750730View attachment 3750729

I like the second one


Wondering, should we make a request thread?

 

[SomethingProblematic]

Made a shit thread so this thread gets less shitted up with requests and stays as a showcase/discussion.

https://kiwifarms.st/threads/ai-image-requests.130976/

 

[Male Idiot]

Do it with the non anime generator thing


I like the second one


Wondering, should we make a request thread?

Where do you set which generator it uses? I can't find it under settings.

Yes, perhaps that would be a good idea. Would you like it if I set it to my avatar?

 

[Sage In All Fields]

But it can't do full comics yet

It probably wouldn't be very difficult to make an AI comic generator. The parts are already there.
All you'd need is:
- Comic layout templates to decide the proportions for each panel you want
- GPT-3 to write storylines based on prompts
- Some glue code to pump GPT-3 into Stable Diffusion to generate each panel

 

[SomethingProblematic]

Where do you set which generator it uses? I can't find it under settings.

Yes, perhaps that would be a good idea. Would you like it if I set it to my avatar?

For anyone following the voldy guide who doesn't want anime shit, for the love of god download the hugging face model, skip novelAI. Everything else is specifically for various types of 2D porn.

Idk what you mean lol, idc what you do. This guy mentioned something about a non-weeb model.

 

[Slav Power]

https://docs.novelai.net/

That would be for NovelAI which is very anime based. But what about Stable Diffusion v1.4 checkpoint? Any tips on what prompts I should use to improve the end result, how I should form my prompts to get as close to the desired result as possible, or generally what settings to use, like steps, CFG scale and all that? Also how do I add tags to the WebUI?

 

[cheese burger69]

Spoiler: Image Dump, 1 collage = same prompt but different seeds (sfw, just very big)

 

[SomethingProblematic]

Spoiler: Image Dump, 1 collage = same prompt but different seeds (sfw, just very big)

View attachment 3750745View attachment 3750747View attachment 3750748View attachment 3750750View attachment 3750753View attachment 3750754View attachment 3750755View attachment 3750757View attachment 3750758View attachment 3750759View attachment 3750760View attachment 3750761View attachment 3750763View attachment 3750764View attachment 3750766

Honestly this might get me back into d&d, would be so much easier to set up scenes. Thanks chuckificial sneedtelligence.

 

[NoReturn]

Do it with the non anime generator thing
I like the second one
Wondering, should we make a request thread?

 

[GumballFapper]

The furries made a google collab notebook for stablediffusion if you have shitty hardware like me:
https://colab.research.google.com/drive/128k7amGCLNO1JGaZhKl0Pju8X7SCcf8V
All you gotta have is a google account, then run through the blocks one by one, easy peasy and you're set in around 15 minutes or less

The yiff model is really disappointing though, it keeps generating retarded blobs like these:


From fucking around with it for around an hour, I found out that if you mention "vaginal_peneteration" you're more likely to get an actual erotic picture rather than a retarded blob, the face and some of the limbs will still continue to be broken though

Seems like we still have a long time until AIs that can put furry artists out of business

 

[Male Idiot]

I managed to make a suitably derp looking Sonichu. He takes after his father.

 

[Road Work Ahead]

So this thing is pretty good at fantasy castles. I can see it getting plenty of use for generating things when I run dungeons and dragons campaigns

 

[Pro. Memer]

I'm using the Automatic1111 frontend. How do I know that I successfully installed the NovelAI module?

Somewhere there's a setup troubleshooting guide, I'm not gonna look for it because I'm lazy, but if you can replicate the sample image, you did it right.

 

[Pope Urban II]

ahh man i love this toy > the noob gui version that is

 

[Bananadana]

Sunrise, Canyon, River.

Some more Sonichu oc. Zap to the Extreme!

Where do you set which generator it uses? I can't find it under settings.

Yes, perhaps that would be a good idea. Would you like it if I set it to my avatar?

The generators are different files that are available here.
This is the normal model/generator "huggingface".

You need to copy the .ckpt file to stable-diffusion-webui/models/Stable-diffusion after downloading.
You can select between multiple .ckpt files in this folder with the "Stable Diffusion Checkpoint" drop-down menu at the top of the WebUI.
(does not even require a restart, click the little circle-spinny thingy to the right of the menu and it should show up.)

 

[Kendall Motor Oil]

FYI:
The novelAI torrent is ~56GB.

 

[lolcow yoghurt]

It turned out cuter than I expected but it's still a bit scary. Thank you :3.

 

[umami's milk]

FYI:
The novelAI torrent is ~56GB.

Jesus. What's taking up so much space? It's gotta be the pictures it pulls data from.

 

[Stalphos Johnson]

Jesus. What's taking up so much space? It's gotta be the pictures it pulls data from.

Most of that is redundant models. The final or pruned models are the most refined and tend to produce the best results. The final model is only about 3 gb which is much more managable. Your torrent client should be able to only pull the files you need. Check the first page of the thread for instructions

 

[cerulean hydra]

I wanted to join in the Roon posting and actually generate her in a nazi uniform, but both waifu diffusion and the novelAI anime model seem to have little to no knowledge of shipgirls.

So anybody crying "this will replace bottom-tier deviantart/twitter/waifu artists", you need not fear. Just pivot to drawing shipgirls, or flavor of the month gacha waifus.