[X] My Assistant

Hijacked the thread to make a generic one, because it's easier for me to clean up when it inevitably gets "ai is shit11!!1" that we've heard 1000 times already taking them off topic. I don't particularly like it, I'm sure most of you reading this don't particularly like it, but you don't need to let it be know here, it's expected and no one gives a flying fuck.

For general tech discussion on anything "AI", including but not limited to, image generation, large language models, decensoring, translating, etc, and the technologies behind them. Not a thread for posting your work, but a thread for asking how to's or general discussion of the technology.

If you don't like AI, feel free to not post unless it's actually helping. Off topic posts will be removed.

I have seen some galleries with images like this one:
https://e-hentai.org/g/2945167/2b71b17db0/

What tools are used to create them?

Tools?
or do you mean Check point or Lora , or something like that?

Looks like generic stable diffusion with whatever 2.5d checkpoint.

If I think on, I'll be combining ai threads here into one generic one, we don't need a bunch for it.

Hijacked and locked some of the other threads. Read the OP before posting, thanks.

Currently using KoboldCPP and Sillytavern to run LLMs locally for roleplaying. My 6GB of VRAM means I have to run them off CPU/RAM with some GPU offloading. I can't do medium to large sized models, but I can still run most quantized 7B to 9B models at an okay speed. Not fast, but acceptable.

I plan to build a new gaming PC sometime soon with 16GB of VRAM so I can start playing around with medium-sized models and try out Stable Diffusion.

6gb is enough for bog standard stable diffusion setup with <2.0 checkpoints, which is what most the stable diffusion images you see are, though it'll be slow as balls on anything large. Trying to do a 1920x1080 image will need all the low vram options enabled, and take about an hour for a single generation.

No clue on stable diffusion xl, or cascade, etc.

What's your favorite art generation AI?
NAI seems to be the best for anime art currently but SD can make use of Loras. What do you think?

This post has been edited by Shank: Sep 18 2024, 10:06

SD has so much customizability. I've gotten good results for anime art depending on the model you use. civitai has a toooon of models that are perfect for just about every art style

i am new here i am huge fan of this website since 2010 so being while and its my first post...
i dont know how its work i never use forum of my entire life and i am not 20...

i dont know if its the right place to post but i see a lot of excellent work on IA on this website so here my question

i made a video game on python using pygame i try to draw the enemy but i am not really good at this i try to colorize picture and if some of you want ot help it can be great.

i try many ia like "playground" or "openart.ai" but it do the really bad thing so please help me

i am new here i am huge fan of this website since 2010 so being while and its my first post...
i dont know how its work i never use forum of my entire life and i am not 20...

i dont know if its the right place to post but i see a lot of excellent work on IA on this website so here my question

i made a video game on python using pygame i try to draw the enemy but i am not really good at this i try to colorize picture and if some of you want ot help it can be great.

i try many ia like "playground" or "openart.ai" but it do the really bad thing so please help me

That's not exactly a question so much as just a blanket statement of needing help. Which is perfectly fine, but you should probably clarify what exactly you need help with. Do you need guides on how to use certain AI generation tools for your asset creation? Do you need someone to create game assets for you? If so, I wouldn't say this is the right place to post a commission.

What would the form of your desired assistance look like?

AI application or software
Need your help

Hello,
I would like to start creating hentai images in the style of Kase Daiki, because I love it.
But I don't know at all what software or application to use and especially the scripts to put.
If anyone could help me get started in this art, I would be happy.
Thanks in advance for anyone who responds to me.

This post has been edited by Shank: Dec 17 2024, 12:25

It would be great if we can have some guidance on flux / stable diffusion.
I am trying to make something with Stable Diffusion 3.5 Medium from Huggingface with Diffusers library but the result is censored.

Here, can I ask which program is used to create the hottest images we have here?

<@Kagoraphobia> tag by tag ID
<@Kagoraphobia> i wanna add 1001
<@Shank> Just checked what tag that was. Nice memorable tag id
<+nasu> that's the binary for 9 which also happens to be the first gallery
<+nasu> everything is connected

I tried to deploy the gemma3:12b model locally and bypass the restrictions with some prompts, and then sent the first draft to my friends who understand Japanese for revision, which seems to increase the translation speed, but most of the limit-breaking tips circulating on the Internet are not suitable for manga translation, can anyone give some advice?

I tried to deploy the gemma3:12b model locally and bypass the restrictions with some prompts, and then sent the first draft to my friends who understand Japanese for revision, which seems to increase the translation speed, but most of the limit-breaking tips circulating on the Internet are not suitable for manga translation, can anyone give some advice?

Any particular reason for gemma besides the language proficiency?

Whats your local llm client?
Whats your jailbreak prompt?

Any particular reason for gemma besides the language proficiency?

Whats your local llm client?
Whats your jailbreak prompt?

I am using gemma3 because other models don't perform very well (they lack careful quantization), and they actually perform quite poorly. As for the LLM client, I was actually using BallonsTranslator, but I have now abandoned this approach because the translator believes it increases his workload and decreases translation quality.

Did this thread fall off the main page? Generally we want to put the AI conversation in here.

Hey Siri, stick this response below into the right context
"Worry less about quantization which will reduce accuracy and more about getting more ram and vram. Running models on a potato generates more potato"