Help with Stable Diffusion image generation

[Evocatus]

I recently started playing with Stable Diffusion, I have no idea what Im doing basically, Im still happy that thing actually works (well kinda, having AMD card does not really help). Anyway Im looking if someone who knows and is willing to help me configure it for 2D cartoon like images, that dont suck obviously.

Would appriciate any kind of pointing to the right direction.

 

[neddihseil]

backend sd? or are you using A11111?

 

[Evocatus]

A1111, but nothing is set in stone, if its better to change to something else im open to that.

 

[neddihseil]

A1111, but nothing is set in stone, if its better to change to something else im open to that.

I would not go backend lol, A1111 is great, and there is plenty of help online for using it. Alas, I use the backend command line, so can't help you here

 

[Evocatus]

OK, thank you anyway. Im actively looking around to try to make it work as I wish. I will make a tutorial thread after its done (if ever).

 

[Zombiedaddy]

I recently started playing with Stable Diffusion, I have no idea what Im doing basically, Im still happy that thing actually works (well kinda, having AMD card does not really help). Anyway Im looking if someone who knows and is willing to help me configure it for 2D cartoon like images, that dont suck obviously.

Would appriciate any kind of pointing to the right direction.

I assume you know of

You must be registered to see links

? all you need is there.

 

[Evocatus]

Yes

I assume you know of

You must be registered to see links

? all you need is there.

Yes, thanks. That seems to be the best source of assets and models.

 

[Hobobobb]

Some advice would be to learn how to weight your prompts, and also keep the seed # in mind.

Prompt weighting is very important: Girl, (Red hair, twin tails hairstyle), (freckles:0.3), eating a burger

Will have much less freckles than just "red head girl with freckles eating a burger"

Seed # I suggest you use Random, until you find something that has the basic elements you like, then stick to that seed for refining your image.

 

[-Sikishi-]

There is also this section on CiviAI:

You must be registered to see links

You can click on the images and see everything used. models, prompts, loras etc. It's a really great way to learn the how tos

 

[azr43l]

I'm a bit late, but look into ControlNet as well. It allows you to have a finer control over poses and other things.

Can't give any detailed tips because I haven't been tracking the latest developments for a few months now (and AI stuff tends to change a lot lol), but just search for tutorials about it and you will find plenty of stuff.