I figured that some of you might enjoy the various stuff I generated between two series (either for testing/practicing purpose or for fun) (follow me on X https://x.com/Nonimousmate to not miss this kind of content)

Hey, what is the method you are using while generating theese pictures ? I mean you are using online AI generator or you downloaded AI yourself and trained it ? Can you direct me to some kind of tutorial ?

Base +6, Whatpumpkin -8, albertzeb -6, Spyromancy +6, Mr. Xecret +6

Hi Arswark,

I'm using local image generation (open sourced AI running on my pc), I intend to do a full tutorial one day .

In the meantime you can follow Fluffscaler's tutorials they are very good:
https://rentry.org/fluffscaler

Here are the core concept that you'll see on the tutorial:

1- You need a good enough GPU (8gb of vram is enough, you can do things with lower but it will limit your options a lot)

2- Choose one of the many UI application (I'm using ComfyUI wich is one of the more complexe to learn but offer more options)
to begin I'd say that Fooocus is probably the best,
you can also check invokeAI or forge/ webUI (forge is the experimental version of the later)
(the is also Krita which is a drawing app that have a plugin to run local AI)
=> if you wanna test multiple UI app "stability matrix" allow to install and manage mutliple UI easilly

3- Choose a model (the model is an AI that was trained to generate image, depending on the training it will be better for some usage and worse for other)
There is few "base" model that were provided by big actors such as stable diffusion and many finetuned or merge model that are mostly made by the community.
Finetuned are base model that got a second training to expand their capabilities and merge are a mix of different model (to get a third one the have a bit of both capabilities)

for nsfw Illustrious/noobai are currently very popular, personnaly I use mostly Nova Furry which is based on noobai itself based on Illustrious itself based on the SDXL base model.
you can find the model (also called checkpoints) on civitai https://civitai.com/models

4- Do some image generations, learn to prompt for your choosen model check online ressources (youtube, civitai articles)
On civitai each model has a gallery of image that were generated using this model, when you click them some have the parameter and prompt that were used to get this result you can copy them and try to get similar result.
(most of mine don't because I do multiple passes so putting a prompt and params would be misleading)
DO NOT EXPECT perfect result from the start, also aim for something simple at first: a single subject and a background for exemple. (multiple characters are hard for AI)

5- Learn to inpaint (redraw some area of an image) upscale (add details to an image) and use lora (they are like model add-on, there are lora for character/pose/style/... available on civitai)

If you have more questions feel free to ask on civitai were I'm most active and like to talk about local generation.
https://civitai.com/user/nonymousmate

Last edited on 27 March 2025, 23:23.

Base +3, Whatpumpkin -8, Mr. Xecret +6, albertzeb -6, Spyromancy +6, Burnout745 +6

Thank you very much, going to try it. let me know when tutorial will be out.

Base +6, albertzeb -6, Mr. Xecret +6