stable-diffusion / AI - PART TWO

[Johnny Cage]

Just messing around a bit more with PerfectDeliberate. I would say that the model tends to produce nice results, but also on average stuff that feels a bit more artificial than most of the other checkpoints I have been using.

[SynchroDub]

Just a little announcement.

While I like what i'm seeing here, so far, please be careful NOT to post any photos where the girl/woman's face has the look of a child.
If a AI-generated woman look A LOT like a young underage girl, please don't post it.

I already saw that some people, outside of this space, already started creating children in bikinis.
Please don't post any of that crap, here.

Same rules for underage content also applies to AI-generated images as well.

Thank you.

[Johnny Cage]

My latest batch was using the "combine faces" trick again.

IIRC, two of those are combinations of Scarlett Leithold and Madi Teeuws.
The other two of Scarlett Leithold and Hannah Ferguson.

[BoInk2]

Originally Posted by Johnny Cage

My latest batch was using the "combine faces" trick again.

IIRC, two of those are combinations of Scarlett Leithold and Madi Teeuws.
The other two of Scarlett Leithold and Hannah Ferguson.

[Johnny Cage]

Originally Posted by BoInk2

It's where you add the names of two different models to the prompt and let SD make like a mix of their facial features.

[BoInk2]

Originally Posted by Johnny Cage

It's where you add the names of two different models to the prompt and let SD make like a mix of their facial features.

Yeah, I thought you'd post some pics.

[Johnny Cage]

Originally Posted by BoInk2

Yeah, I thought you'd post some pics.

Not really been doing that much interesting stuff, mostly trying to figure out prompts that give more satisfactory results in PerfectDeliberate.

[maxhitman]

Originally Posted by Johnny Cage

Just messing around a bit more with PerfectDeliberate. I would say that the model tends to produce nice results, but also on average stuff that feels a bit more artificial than most of the other checkpoints I have been using.

Cool pictures.
The women look very sexy, but some of these "checkpoints" ,which are not
meant for doing photo-realistic pictures, do tend to make the female face
to look almost like those old DAZ-3D models I used to see. Specially the
old "Genesis 2" models where all the females look the same and have
doll-like features.

If you test the Checkpoints by using a pre-trained T.I. or Lora of a
famous celeb face, it tends to distort the face to look more like a doll or
a very cute cartoon female. Usually never the same exact face as the
real person.
In other tests i have done, if you use a T.I. , or Lora, of an older celeb face,
it tends to make them look younger - which sometimes is cool.

Personally, I like to stay away from using them, unless I want to create some
type of artsy portrait of someone, but some times I use them as one
of the "Checkpoint-Mergers" - in the B or C - selection just so the A.I. can
create a smooother skin texture and it also helps to create better fingers.
With some cartoon checkpoints, or 2.5d, or 3D type of art Checkpoints,
it does help a little to create better hands.

It´s as if the A.I. is better tuned to create better fingers in "Art" rather than
if the images are for Photo-Realism.

Anyway,
at the moment I am still testing words and using a Thesaurus to find
other interesting words, to use in my Prompts.
I am still sticking to my Special Prompt, which for me, is the best ever for
creating photo-realistic pictures. Each section of the prompt is separated by
a comma, and each "section" (as I call them) triggers a response
from the A.I. to create exactly what I want to see - Without the use of
any special pre-trained Lora, Lycoris or Textual-Inversion (T.I.)
I just change the words in some sections to see what happens with
the A.I..
In many cases I discovered that it does create ALOT of things, which I
often see being posted at Civitai .
I guess most people are still not aware of how to create certain things,
and thus, they waste their time creating a Lora or T.I. to make their A.I. do
things for them.
The other day, I saw someone post a Lora for creating "long hair. "
I said to myself, Huh?
Seriously ?
You need a Lora for making "long Hair" on your female or male model ?
Did you know that your A.I. can create just about 80% of all hair styles
and hair colors that ever existed ?!
Just write it in your prompts.

[maxhitman]

Here is a Very Interesting article which Everyone using Stable Diffusion or any other
A.I. program should read and see....

Exploring 12 Million of the 2.3 Billion Images Used to
Train Stable Diffusion’s Image Generator


One of the biggest frustrations of text-to-image generation AI models is that they feel like a black box. We know they were trained on images pulled from the web, but which ones? As an artist or photographer, an obvious question is whether your work was used to train the AI model, but this is surprisingly hard to answer.

Sometimes, the data isn’t available at all: OpenAI has said it’s trained DALL-E 2 on hundreds of millions of captioned images, but hasn’t released the proprietary data. By contrast, the team behind Stable Diffusion have been very transparent about how their model is trained. Since it was released publicly last week, Stable Diffusion has exploded in popularity, in large part because of its free and permissive licensing, already incorporated into the new Midjourney beta, NightCafe, and Stability AI’s own DreamStudio app, as well as for use on your own computer.

But Stable Diffusion’s training datasets are impossible for most people to download, let alone search, with metadata for millions (or billions!) of images stored in obscure file formats in large multipart archives.

So, with the help of my friend Simon Willison, we grabbed the data for over 12 million images used to train Stable Diffusion, and used his Datasette project to make a data browser for you to explore and search it yourself. Note that this is only a small subset of the total training data: about 2% of the 600 million images used to train the most recent three checkpoints, and only 0.5% of the 2.3 billion images that it was first trained on.

read more and see the images at...

http://waxy.org/2022/08/exploring-12-million-of-the-images-used-to-train-stable-diffusions-image-generator/

In the webpage, they even give out some of the links to some of these data-sets
of images which were used to train the A.I. - Billions of images !
I suggest to be careful of what you want to "download" because you might
end up blowing up your computer with too much data-information


Some of these 6+ Billion "tagged" images are still stored and
can be seen at...
LAION-5B: A NEW ERA OF OPEN LARGE-SCALE MULTI-MODAL DATASETS

http://laion.ai/blog/laion-5b/

[maxhitman]

Some people ask me,
how many celebrity faces are inside Stable Diffusion ? , and can it create them all?
It does create some of them VERY GOOD, while some other people you may need to
keep prompting your A.I. to create a few more images, until the A.I. starts to create
their faces very good. What I mean by this is, keep using the same prompt asking to
see a particular known face and create over 50+ images. Usually after the 40-50
images created the A.I. will begin making the faces very well.

So,
HOW MANY Famous Faces are there inside Stable-Diffusion ?

Here is the complete list...
Each Page has 100 names ... and there are 1,996 Celebrities to read all the names
In the pages you can also see How many images of each person are used for A.I. training.

http://laion-aesthetic.datasette.io/laion-aesthetic-6pls/celebrities