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

[Post Reply]

you're telling me you wouldn't tap this?

Spoiler: one foxy mama

View attachment 6547443

Spoiler: they did look slightly similar though

View attachment 6547455View attachment 6547460

Delete this before The Last Stand starts another goon sesh.

 

[Saint Agustin]

Can anyone of you suggest me how to get buzz on civitai? Nigger faggots are fucking hard and going the crypto route and I just want to make anime tits without having to pay.

 

[Squishie PP]

Can anyone of you suggest me how to get buzz on civitai? Nigger faggots are fucking hard and going the crypto route and I just want to make anime tits without having to pay.

Get any GPU with >4GB VRAM from the last 20 years and do it on your own machine.

 

[Baraadmirer]

Can anyone of you suggest me how to get buzz on civitai? Nigger faggots are fucking hard and going the crypto route and I just want to make anime tits without having to pay.

What Squishie said. Literally 99% of what I do is locally on my own device. You can upload your anime tits online afterwards to get buzz that way.

 

[Squishie PP]

What Squishie said. Literally 99% of what I do is locally on my own device. You can upload your anime tits online afterwards to get buzz that way.

<snip>

Will advise using NVIDIA GPUs with Pascal architecture or any of their newer architectures. Otherwise the tools you use will be a lot more difficult to set up.

 

[macrodegenerate]

You can run AI locally on your phone now. The models are as trash as you expect though. It's 3.3 3B Q6

 

[Saint Agustin]

Will advise using NVIDIA GPUs with Pascal architecture or any of their newer architectures. Otherwise the tools you use will be a lot more difficult to set up.

My laptop is old and can only handle doing voice speech alteration using my GPU and I don't have the patience to set up GitHub repositories...

I wonder though, doing realistic pictures (real woman and not anime tits) does it takes more processing power???

 

[Squishie PP]

My laptop is old and can only handle doing voice speech alteration using my GPU and I don't have the patience to set up GitHub repositories...

I wonder though, doing realistic pictures (real woman and not anime tits) does it takes more processing power???

Both take roughly the same amount of processing power. It depends on which model you are using and whether or not the weights have been pruned, but if only the prompt differs then it will be the same.

If all you want to do is proompt and don't care about the autistic comfyui shit, Jellybox will be a lot easier to use on your own machine: https://jellybox.com/ It doesn't require setting up git or python environments like many of the other tools do. All you do is install. So basically as close to local AI model inference for niggers as you can get.

If you don't use Windows or Mac, check back with me in a while. I have something similar for Linux and maybe BSD in the works.

 

[Baraadmirer]

My laptop is old and can only handle doing voice speech alteration using my GPU and I don't have the patience to set up GitHub repositories...

I wonder though, doing realistic pictures (real woman and not anime tits) does it takes more processing power???

I haven't noticed any appreciable difference; I use different checkpoints to help generate different styles. What might affect processing time, besides features like the size of the generated images and hires fix is the sampler you're using. Euler a is one of the faster ones.

 

[frozen_runner]

Gentlemen, the only way this could be better is if the U.N. logo were in the background and the people had rifles.

Spoiler: 🇮🇳 NSFW INDIAN CONTENT 💩

 

[The Mass Shooter Ron Soye]

Google offers its AI watermarking tech as free open source toolkit (archive)

Google uses a version of SynthID to watermark audio, video, and images generated by its multimodal AI systems, with differing techniques that are explained briefly in this video. But in a new paper published in Nature, Google researchers go into detail on how the SynthID process embeds an unseen watermark in the text-based output of its Gemini model.

Fuck yo watermarks and GayyyyI, nigga.

 

[Gog & Magog]

Google offers its AI watermarking tech as free open source toolkit (archive)

Fuck yo watermarks and GayyyyI, nigga.

What if someone screenshots it?

 

[La Flanaderia de Flan]

I was bored

Spoiler

 

[Starved Artist]

Can anyone of you suggest me how to get buzz on civitai? Nigger faggots are fucking hard and going the crypto route and I just want to make anime tits without having to pay.

Download WebUi-Forge. Generate something Uber Kawaii, something Risque, then something NSFW that doesn't have sameface and upload it to the gallery.
Try to figure out what this week's gooning obsession is and make SFW versions of it and you'll get collateral buzz.

 

[macrodegenerate]

So you can run Flux.dev locally on anything with 12GB now. Someone figured out how to make and implement .gguff quants. I'm running Q8 on the text encoder and the actual model. It's still going to take some work to refine it, but the results aren't bad so far.

 

[Baraadmirer]

Download WebUi-Forge. Generate something Uber Kawaii, something Risque, then something NSFW that doesn't have sameface and upload it to the gallery.
Try to figure out what this week's gooning obsession is and make SFW versions of it and you'll get collateral buzz.

Also make sure you're using a current popular checkpoint. Today's featured images all appear to be made with FLUX.

 

[Darkholme's Dungeon]

https://github.com/erew123/alltalk_tts/

You can use this to do voice cloning from a short snipped of audio, but the results are very hit or miss and the quality is kind of low. But what you could do is feed the results from this into RVC (https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/main/docs/en/README.en.md) and use one of the easily trained and better sounding RVC voice models overtop the generated voice. You don't need to use alltalk, you could use any TTS as a base. I think this is how all those AI meme videos are made, or rather, the ones that don't use ElevenLabs.

I think ElevenLabs will probably still have better quality out of the box, and is easier to use, but you have to pay them and give them your personal information. Local is always better.

Have you tried V2 alltallk? Just curious as to if there's a noticable difference (good or bad).

 

[Soggy Floppa]

Have you tried V2 alltallk? Just curious as to if there's a noticable difference (good or bad).

I haven't

 

[AnimeSucksCopeAndSneed]

At this point the two have really diverged. SDXL is really the best for SFW. While PonyXL is the best for NSFW easily. You are still going to have problems with text.

Contentious opinion honestly.

 

[macrodegenerate]

Contentious opinion honestly.

You use a checkpoint(YiffyMix) based off of a distilled model of SDXL(Lightning). Pony XL might actually improve your art if you give it a shot. The Lora diversity is impressive on civit.ai.