Stable diffusion thread

[doogyhatts]

looks creepy

When she is fed with bottle gourd juice, her skin will change to a sickly green.

 

[doogyhatts]

just use reactor for the face can liao.
this one so ma fun.

and somemore, the body type is inconsistent.

ReActor uses insightface, which trains for front view, not side view.
As for consistent body type, I would generate it separately in a grid and then combine it manually with the specific head.

 

[doogyhatts]

Haiper now can only do 10 free creations per day.
This will impact those depending on it solely to animate their images.

 

[doogyhatts]

i want to a full animated one.

This music channel is monetised, while many others failed to do so.
The art is AI generated, but the music may not be.

 

[AikiBoy]

This music channel is monetised, while many others failed to do so.
The art is AI generated, but the music may not be.

I want to do that style of render... any tips?

 

[AikiBoy]

Haiper now can only do 10 free creations per day.
This will impact those depending on it solely to animate their images.

to use haiper for youtube monetization need to buy for $24 sub.. zzz

 

[doogyhatts]

I want to do that style of render... any tips?

Use any of the anime models such as AnimagineXL.

 

[AikiBoy]

Use any of the anime models such as AnimagineXL.

Install in stable diffusion?

 

[doogyhatts]

Install in stable diffusion?

Install Fooocus using Stability Matrix.
YT have tutorials on using Fooocus.

Fooocus is like Midjourney but only for SDXL models.
It comes with a prompt comprehension system better than CLIP.
The github page for Fooocus has a lot of other information too.

 

[AikiBoy]

Install Fooocus using Stability Matrix.

so many app...

 

[doogyhatts]

so many app...

Install Fooocus using Stability Matrix.
YT have tutorials on using Fooocus.

Fooocus is like Midjourney but only for SDXL models.
It comes with a prompt comprehension system better than CLIP.
The github page for Fooocus has a lot of other information too.

@AikiBoy
Try using Fooocus before moving onto ComfyUI.

 

[doogyhatts]

Krita now has regional prompting.

 

[AikiBoy]

Krita now has regional prompting.

wao liao... endless update catch no ball

 

[zheng]

Krita now has regional prompting.

cool bery nice feature

 

[doogyhatts]

wao liao... endless update catch no ball

You don't have to upgrade every time new stuff comes out.
Come back to it when you actually need to use it.

Regional prompting is mainly used to generate multiple characters in the same scene, in which each character has its own region of space within the image.

 

[doogyhatts]

SD3 medium released at HF.

 

[doogyhatts]

Raw output from PixArt-Sigma, upscaled with photon model.
It is a Kappa Musume that shapeshifted into human form.

Comparing with SD3 medium.
Looks like I will have to stick with PixArt-Sigma to generate my horror-based content.

 

[doogyhatts]

Only paying for Stable Assistant (chatbot) or Stable Artisan (discord), then can get good image output.

 

[doogyhatts]

Lumina T2X uses DiT architecture.
Can do different outputs, not just image.

 

[doogyhatts]

More comparisons.