Saint-Archives.com Forum

Members Login
Username 
 
Password 
    Remember Me  
Post Info TOPIC: AI-Generated Art


Senior Member

Status: Offline
Posts: 112
Date:
AI-Generated Art
Permalink  
 


This is my first attempt at using Stable Diffusion.

There are a lot of the same type of pictures, but it was this style that turned out best for me.

The original size is 512x512px. After that I used Topaz Gigapixel for 2x upscaling.

 

https://pixhost.to/gallery/lqmQP

 

 

 


-- Edited by sam917 on Wednesday 28th of June 2023 10:53:25 PM

Attachments
__________________


Moderator

Status: Offline
Posts: 1092
Date:
Permalink  
 

Interesting. Nice work.

Is this merging a photo of Silvia and a photo of e.g.Taylor Swift, or is it taking a photo of Silvia and using some AI model to make her look like Taylor Swift (or vice versa)?

As with all AI the results are mixed. Mostly quite nice, but the "Stormy Daniels" ones don't work well.

If you number them in order presented in your post, I'd say the best ones are nos. 14, 16, 17, 18 and especially 19. I guess that's possibly because the target face shape is closer to Silvia's own face shape. Not sure who the target is though.

 



__________________

Spike



Senior Member

Status: Offline
Posts: 112
Date:
Permalink  
 

To get started, I downloaded the basic Stable Diffusion model (Checkpoint model).


After that, I used 8 photos of Silvia from one photoset to train the neural network. The more photos the better, but it will take longer. As a result, I got Silvia's LoRA model. LoRA is patch files to checkpoint models for modifying styles.


After that I used Silvia's LoRA model and LoRA model with cute Asian faces (50% power). As a result, I got about half of the pictures that you see. The second half is experimenting with other settings.


To generate a picture, you need to use prompts, for example: realistic,(lips closed)1girl, eyewear on head, solo, breasts, looking at viewer, sunglasses, jewelry, earrings, white hair, lips, upper body, hair ornament , hair between eyes, bare shoulders, red lips, large breasts, ponytail, neon lights, closed mouth, bangs.

I took photos from this photoshoot.

 


-- Edited by sam917 on Thursday 29th of June 2023 11:30:43 AM

Attachments
__________________


Senior Member

Status: Offline
Posts: 112
Date:
Permalink  
 

At the moment for the difficulty is to create a good Silvia's LoRA model. It need to experiment a lot.

__________________


Moderator

Status: Offline
Posts: 1092
Date:
Permalink  
 

That's pretty impressive from a model generated from just 8 photos!

 



__________________

Spike



Senior Member

Status: Offline
Posts: 112
Date:
Permalink  
 

After that, I used 14 photos from another photoshoot, but the result was very bad. So far, I can't figure out how to choose the right photos. There are many nuances. But the possibilities of neural networks are amazing.



__________________


Senior Member

Status: Offline
Posts: 112
Date:
Permalink  
 

New styles. What do you think?

 


__________________


Member

Status: Offline
Posts: 9
Date:
Permalink  
 

At last! I've been waiting for someone to start producing Ai pics of Silvia. The closest we will get to new pics. Will have a close look at these later but thank you doing this.

__________________


Member

Status: Offline
Posts: 9
Date:
Permalink  
 

Second set were a big improvement. There were a few in the first set that were a bit like Taylor Swift and even Amanda Holden (a UK tv presenter)

__________________


Veteran Member

Status: Offline
Posts: 62
Date:
Permalink  
 

U think u'll be able to produce nsfw images in the future? 



__________________


Senior Member

Status: Offline
Posts: 112
Date:
Permalink  
 



__________________


Moderator

Status: Offline
Posts: 1092
Date:
Permalink  
 

I like where you're going with the fantasy-style images, Sam.

Still some work to do on refining the face model, and I'd have to say the AI-generated nipples and areolae are rather ugly. Probably better to make her fully clothed if that's the best the AI can do.

 



__________________

Spike



Veteran Member

Status: Offline
Posts: 62
Date:
Permalink  
 

These are really cool as far as the AI works! I really like the design of the 3rd one specially. Pls go on with ur creations until u can master the production of these pictures, because it's really entertaining 



__________________


Senior Member

Status: Offline
Posts: 112
Date:
Permalink  
 



__________________


Veteran Member

Status: Offline
Posts: 62
Date:
Permalink  
 

Hey! These are really beautiful! My favorite outfit must be the one from number 2, but I love the outdoor composition of 5 and 6! Also it's really nice to see images of Silvia with a more Victorian Goth style this time around.

Keep the fantastic work coming! This is really inspiring!



__________________


Senior Member

Status: Offline
Posts: 202
Date:
Permalink  
 

 Sans titre 1 - Copie.bmp

Sans titre 4.jpg



-- Edited by kath78 on Wednesday 16th of August 2023 12:32:48 AM

Attachments
__________________


Senior Member

Status: Offline
Posts: 112
Date:
Permalink  
 



__________________


Veteran Member

Status: Offline
Posts: 62
Date:
Permalink  
 

These are really cool. Please keep the great work, it's so nice to see Silvia's face in this fictional pictures with different styles 



__________________


Senior Member

Status: Offline
Posts: 202
Date:
Permalink  
 

     


 





 


 


 


 



-- Edited by kath78 on Wednesday 15th of November 2023 10:41:54 PM

Attachments
__________________


Senior Member

Status: Offline
Posts: 202
Date:
Permalink  
 

45008b8215a895907bf4f5fdebf74dcf_768_768.jpg7832a8faf53d2333f742e7c781c86655_768_768.jpg

 

 



-- Edited by kath78 on Monday 29th of January 2024 01:47:50 AM

Attachments
__________________
1 2  >  Last»  | Page of 2  sorted by
 
Quick Reply

Please log in to post quick replies.

Tweet this page Post to Digg Post to Del.icio.us


Create your own FREE Forum
Report Abuse
Powered by ActiveBoard