stable-diffusion / AI - PART TWO

[maxhitman]

Originally Posted by MaxLord

I think I'm going to look into creating a few of my own Loras as well. What is the best tutorial to look into for that?

from my post # 45


I learned more from these videos than from watching many others.
On other YouTube videos they always skip some important steps.

These are the 3 videos you must watch...

step 1 - installing the Kohya_ss for creating Lora
by Olivio Sarikas  (His video tutorials are excellent!)
LORA: Install Guide and Super-High Quality Training 
http://youtube.com/watch?v=9MT1n97ITaE&t=399s


Creating a LoRA weight using kohya_ss GUI, 
part 1: Dataset creation and preparation 
- He is making a LORA of Jennifer Aniston 
http://youtube.com/watch?v=N4_-fB62Hwk&t=148s

Creating a LoRA weight using kohya_ss GUI, 
part 2: Training the model and creating outputs
http://youtube.com/watch?v=k5imq01uvUY

Keep in mind that when you install this "Lora+Embedding" making gadget
you will have to install a whole new "stable-Diffusion" , which will go
into your Documents file section of your C-Drive.
So you will have one version for Creating images, and another special
version which is just for creating the "Loras".
Watch the videos carefully and follow their instructions.


I often record these YouTube videos for my own Tutorial files
for later viewing, because some times the videos get deleted
by YouTube.
So save them !
If you download and use 4K-video Downloader (it is FREE),
http://4kdownload.com/38
you will be able to save any YouTube videos (in High-Resolution)
to your computer and watch them later, even if your internet
connection fails.

I wish you good luck making Loras
I am already making my 5th one.
It is easy to make but a little time consuming just to get the proper
HQ images, specially if you are doing old classic film stars.
Making the new celebrities is easy , because there are already tons of
HQ+HD images available to use for making LORAs faster.

[Johnny Cage]

Originally Posted by pelham456

who is the one in post 39 pls? is that based on a real celeb or a fully-created ficticious one?

HUBBA HUBBA!

nata lee also really nice, tho not quite realistic. tbh, she's better than the real one!

You mean this one?

Originally Posted by Johnny Cage

Messing around a bit with nighttime lighting in the background.

That isn't exactly based on a celeb, though it kind of is as well.
That was also around the time I was trying out combining facial features of two models.
This was a prompt that asked for a girl that combined Scarlett Leithold and Colleen Cole (both models).

If you are not familiar, this is what they look like. So the face of this SD creation is meant to be a mix of the two.

Scarlett Leithold


Colleen Cole


*edit*

Decided to run the same prompt again, just for fun. This is what I got this time around:

[Johnny Cage]

So expanding a bit on the face-mixing thing. I mostly only do this with models, less so with actors. I find that any really iconic actor, the mix gives very stylized results that don't look very natural.

To give an example, this mix of Raquel Welch and Sophia Loren.



You can recognize facial traits from both but the end result also looks somewhat unnatural.

To stay in the 1960s theme, Julie Newmar and Ann-Margret:



Slightly better but still too stylized.

By contrast, the Anne Hathaway & Marisa Tomei hybrid looks more like a real person.



I'd also add that it's important to try and find facial combinations that might work well together. e.g. here is one I tried of Emily Blunt and Elizabeth Olsen, but I find the combo does not really work that well, even though both women are beautiful on their own.

[maxhitman]

Dominika C

[BoInk2]

What kind of hardware do you guys run this on?

[maxhitman]

Originally Posted by BoInk2

What kind of hardware do you guys run this on?

This is an A.i. program called Stable Diffusion version 1.5, which can create
nude images and art renderings. There is already a version 2.0, but it
has this NSFW/Nudes ability removed from its source code.
So, if you want to create something with a "nipple" showing in your
pictures, you have to use version 1.5.
It is used for home computers, so you can create ANY image without
internet connection. Using a laptop is possible, but you have to read some
online tutorials how to do that, although the A.I. works best on a home PC.

As for computer specs, you really don´t need a top and expensive computer.
I am using a 10 year old computer and it does the job great. Many parts on
it date back 15+ years, but it works ! LOL

One important thing you DO NEED is a good graphics card.
The single most critical component for Stable Diffusion is your graphics card (GPU).
A NVIDIA brand is highly reccomended.
The best are these NVidia ....
RTX 3060, RTX 3060 Ti, RTX 3070, RTX 3070 Ti, RTX 3080, RTX 3080 (12GB),
RTX 3080 Ti, RTX 3090, or RTX 3090 Ti
...these are more expensive....RTX 4090, RTX 4080, and future 40-series GPUs

Minimum Requirements to run Stable Diffusion
These requirements include a GPU with a minimum of 6GB VRAM,
which is found in most modern NVIDIA GPUs. Additionally, approximately 10GB of
storage capacity on either a hard drive or solid-state drive is necessary.

Read this important information to know if you have a computer
able to use this....

Stable Diffusion PC system requirements: what do you need to run it?
By Jon Martindale - February 27, 2023

http://digitaltrends.com/computing/stable-diffusion-pc-system-requirements/

............................................................................
Stable Diffusion PC Hardware Requirement: Detailed Guide
 by decentralizedcreator - March 7, 2023

http://decentralizedcreator.com/stable-diffusion-pc-hardware-requirement-detailed-guide/?utm_content=cmp-true

............................................................................

Search Google using these words...
what type of grapichs needed to use stable diffusion?

............................................................................

VIDEOS...

Stable Diffusion 1.5 - Windows Installation Guide [Tutorial]
by Scott Detweiler ---- BEST Video how to do it properly !!
http://youtube.com/watch?v=ycQJDJ-qNI8

He also posts ALL the Links and necessary things you need to install
S.D. into your computer.

[Johnny Cage]

Some more facial mixing experiments that I thought came out looking alright:

Madison Beer & Ariana Grande



Hannah Ferguson & Hailey Clauson



Nina Agdal & Marloes Horst



And one more that did not work at all, Claudia Schiffer and Heidi Klum:

[maxhitman]

Originally Posted by Johnny Cage

Some more facial mixing experiments that I thought came out looking alright:

Madison Beer & Ariana Grande

....

Hannah Ferguson & Hailey Clauson

....

Nina Agdal & Marloes Horst

...

And one more that did not work at all, Claudia Schiffer and Heidi Klum:

...

Those look pretty cool
You are turning out to be a Doctor Frankenstein ! Mixing up all
those body parts.

I read it some where that if you write in your prompt, for example...

a realistic photo of a (Elephant:0.5, Giraffe:0.5) walking in a field in Africa.

You would get a mixed hybrid "creature" looking part elephant and
part giraffe. LOL
The A.i. will just follow your instructions and create anything you ask.

I guess this trick also works for some fun human mixed-faces.

[Johnny Cage]

A few more hybrid creations I ran.

Katherine McNamara & Brigitte Bardot



Monica Bellucci & Claudia Cardinale



Heather Graham & Sharon Stone



Lais Ribeiro & Adriana Lima

[maxhitman]

"... You are one ugly muthafucker "
(Arnold Schwarzenegger in 1987 film - PREDATOR )
.
The Predator: "... Want some candy ? "

King Willie: " You can't see the eyes of the demon, until him come a-callin' ! "
(1990 film - PREDATOR 2)

------------------------------------------------------------

I have tested many diferent checkpoints in Stable-Diffusion before, and no
matter what useful prompts I used, this A.I. simply does not know what a
"Yautja" is and most often creates a human-like-devil or some other
bizarre creature.
If you use the word "predator" in your prompt, the A.I. thinks its a Lion or
some other type of predator animal. The A.I. simply does not know what this
alien creature looks like.
Therefore,
I created this Lora to teach the A.I. to create this fictional
hunter alien-creature for us to use in our art and pictures.

-

-

-

Download it at Civitai ...

Yautja - The PREDATOR - v.1  " No Helmet " version

http://civitai.com/models/37153/yautja-the-predator-v1-no-helmet

In this version 1, the Predator does not use a helmet.
I will post a version with The Predator using the famous 
helmet in a few days from now.

...and it also creates "Female Predators" LOL

-

-