AI [AI Art] - Show Us Your AI Skill *NO TEENS*

[Hot-HighSchool-GaL]

Not Really Ai Generated But I Used Some Ai Tools To edit the images
View attachment image11.webp
View attachment image.webp

If interested and want to see more of my creation
See my Thread

 

[felldude]

Not Really Ai Generated But I Used Some Ai Tools To edit the images
View attachment 2709951
View attachment 2709952

If interested and want to see more of my creation
See my Thread

Looks like the Anime ERSGAN up-scaing

 

[felldude]

AI reflections and Reflections on AI

So here is an example of 100% text to image and an example of somthing that can be done better in a 3D application



One work around would be to mock up the image in Blender/ Daz etc with and then do an Image to Image.
As long as the noise level was correct the reflection should be maintained but updated in the 3d mock up.

Here is a poor quality 3d render used and the output is closer to what we need...same seed as above.



Just resit the urge to put mirror or refection in your description.

Works on real images also if you don't want or need a certain mock up.
Here is an example with an Instagram by Paulina Porzikov (Again same seed and settings as the first post)

First reflection is correct, subsequent are not, could be in-painted however this is a 2 minute example render so

 

[Hot-HighSchool-GaL]

Wel

Looks like the Anime ERSGAN up-scaing

Well Yeah I used a Run Around Model of ESRGAN
And GFPGAN To Make Face Look Better and some other things too
And the Image You see is drawn by me. Since I am New I have to work my way up

 

[felldude]

Wel

Well Yeah I used a Run Around Model of ESRGAN
And GFPGAN To Make Face Look Better and some other things too
And the Image You see is drawn by me. Since I am New I have to work my way up

Do you prefer GFPGAN over Codeformers when doing anime style

 

[Hot-HighSchool-GaL]

Do you prefer GFPGAN over Codeformers when doing anime style

CodeFormers is Nice for animated,3D, 2D, hentai because it give images artificial depth
On the other hand
GFPGAN is Great For Realism because it gives realistic details i would recommend GFGAN for Surreal, Realistic and other genres which requires detailing.
Well it's my experience
Situations can differs

 

[felldude]

CodeFormers is Nice for animated,3D, 2D, hentai because it give images artificial depth
On the other hand
GFPGAN is Great For Realism because it gives realistic details i would recommend GFGAN for Surreal, Realistic and other genres which requires detailing.
Well it's my experience
Situations can differs

I'll try it out, Stable diffusion webui uses 0.5 codeformers as its default but can be changed to GFPGAN

 

[Hot-HighSchool-GaL]

AI reflections and Reflections on AI

So here is an example of 100% text to image and an example of somthing that can be done better in a 3D application

View attachment 2709954

One work around would be to mock up the image in Blender/ Daz etc with and then do an Image to Image.
As long as the noise level was correct the reflection should be maintained but updated in the 3d mock up.

Here is a poor quality 3d render used and the output is closer to what we need...same seed as above.

View attachment 2709988

Just resit the urge to put mirror or refection in your description.

Works on real images also if you don't want or need a certain mock up.
Here is an example with an Instagram by Paulina Porzikov (Again same seed and settings as the first post)

View attachment 2710029

I'll try it out, Stable diffusion webui uses 0.5 codeformers as its default but can be changed to GFPGAN

Go Ahead.
Don't you think your CodeFormers Version is quite old?
I am not one to say cuz I don't really use ai's but see to it

 

[Hot-HighSchool-GaL]

Go Ahead.
Don't you think your CodeFormers Version is quite old?
I am not one to say cuz I don't really use ai's but see to it

I'll try it out, Stable diffusion webui uses 0.5 codeformers as its default but can be changed to GFPGAN

I'll try it out, Stable diffusion webui uses 0.5 codeformers as its default but can be changed to GFPGAN

And one more thing as for 3D renders i am going to upload some of mine too
On my thread (in about 1-2 hour once i complete it) chech it out
https://f95zone.to/threads/some-fan-arts-i-made-nsfw.164351/

 

[felldude]

Go Ahead.
Don't you think your CodeFormers Version is quite old?
I am not one to say cuz I don't really use ai's but see to it

That is the value it is applied at Codeformer model is codeformer-v0.1.0.pth
GFPGAN is GFPGANv1.4 and has two other models associated with it maybe it isnt default because it is more intensive like using SWIN_IR as your default up-scaler.

Codeformer Left - GFPGAN Right
Both left at 0.5 rate.



Side note GFPGAN sometimes puts facial hair and other details that may not be image appropriate ignoring negatives that might be why codeformer is still used.

 

[Sharinel]

I like this one, at first glance it looks like B&W pencil but then you notice some muted colours

Spoiler: Large pic

You don't have permission to view the spoiler content. Log in or register now.

 

[felldude]

100% text to image with 3 custom LORA's

 

[felldude]

Dark Magican Girl Cosplay

Some of my LORA's are now on

You must be registered to see the links

- If you want to Beta test unreleased content DM me

 

[Ptah]

Hey guys, I didn't knew there was a post about AI in here. I've been tinkering since august of 2022, but I left at 1.4SD because IT was evolving so fast that things get outdated almost every week.

By the way, have you guys found any trick or path to maintain model consistency? Not only on people but also on objects. I'm a total noob on the SD Extras like "impaint" etc.

Here some of my works:



SCI-FI is a BITCH to work with, now some LEWD art:

 

[sharlotte]

Ptah , there's also another thread in here for advice and tips (and more images).

 

[Ptah]

Ptah , there's also another thread in here for advice and tips (and more images).

Thank you so much

 

[Rumbrandt]

Toned biker girl

 

[modine2021]

The Two Zoe's

 

[modine2021]

.

 

[modine2021]

.