Is their a Help threat for AI Art (Stable Diffusion) ?

[Illhoon]

I am especially wondering where i could get infos what things like the Diffrent Sampling Methods , CFG Scale, Seed img2Img, do exactly aswell as help or a collection threat on Diffrent Prompts people use to get what results

 

[DreamingAway]

These can vary between Checkpoints / Embeddings / Hypernetworks .. It really depends on how the data was trained, what bias your prompts will have.

Some great resources on how best to prompt images: (Including sampling & CFG, etc)

You must be registered to see the links

You must be registered to see the links

You must be registered to see the links

Hope this helps.

 

[Illhoon]

These can vary between Checkpoints / Embeddings / Hypernetworks .. It really depends on how the data was trained, what bias your prompts will have.

Some great resources on how best to prompt images: (Including sampling & CFG, etc)

You must be registered to see the links

You must be registered to see the links

You must be registered to see the links

Hope this helps.

Ah thank you Very much.

I got 2 new questions wondering if you could answer them.

What exactly does the model do you select on the top left in the dropdown menu ? is it the artstyle its gonna be ?
and
How do you use Img2Img is there a way that you can input a image thats already made and then let the ai use the faces in a new creation or making a IRL pic into a anime version? or is that not possible?

 

[DreamingAway]

What exactly does the model do you select on the top left in the dropdown menu ? is it the artstyle its gonna be ?

Model checkpoints are basically trained material that interpret prompts and guide denoising.

Some models are general purpose like the stable diffusion ones, and some are very specific to a type of art style or subject matter (Anime, specific artist styles, NSFW content, Photorealistic, etc)

Starting Places to Look for Models

You must be registered to see the links

You must be registered to see the links

You must be registered to see the links

--

How do you use Img2Img is there a way that you can input a image thats already made and then let the ai (sic..) make a IRL pic into a anime version? or is that not possible?

Yes, very possible.

Img2Img allows you to take an image as your starting point and to seed that in your denoise, depending on the denoise strength you use, that picture will inject more and more influence into your output.

This allows you to iterate on an image to seed different art styles, example:
https://f95zone.to/threads/ai-art-show-us-your-ai-skill.138575/post-9681618

(I didn't use img2img for this specific example - I worked off a similar seed / prompt then injected bias from a hypernetwork instead - same concept though)

let the ai use the faces in a new creation

If you want to take an existing picture and only redraw the face, or the body around it, you would want to use inpainting instead - Inpainting does the same as img2img but only in a very specific region of the image that you decide.

--

Does it help?