Stable Diffusion

MadLogiC (SiN)

MadLogiC (SiN)

Well-known member
Featured Contributor
Messages
32
Reactions
607
Points
83
Website
www.patreon.com
Just is just a random post based on my thoughts with Stable Diffusion AI, its frikken awesome :p
I've spent the past few hours generating my model and i gotta say, the realistic output is impressive!
 
hell yeah!

1 (2).jpeg
111.jpeg
222.jpeg
 
Could someone explain what you talking about? I researched it's some sort of an AI image generator, is this correct? Or what the hell is going on in here? What's so special about those pictures you posted?
 
HolySchmidt said:
I researched it's some sort of an AI image generator, is this correct?
Click to expand...
Yes, it's a free & open source one

HolySchmidt said:
What's so special about those pictures you posted?
Click to expand...
they're made from scratch (kind of, long story) by the AI in seconds based on what it was requested of it

here's an example I did recently for a scifi alien magazine. I used an input image as a main guideline and a description like "magazine page, alien, colorful," etc

1 (12).jpeg
1 (15).jpeg
1 (11).jpeg

you can see that if follows a theme but they're all different, you could keep making them for millions of years and they'll keep being unique

or adjust it a bit and go in a different direction

1 (1).jpg
1 (8).jpeg
1 (22).jpeg

infinite artwork at the press of a button essentially. it's gonna make quite the dent
 
Holy shit, that looks amazing. Thanks for the explanation! I would be interested in how the stock image looked like.

But doesn't this kill a whole design industry? Digital artworkers? I did some ptotoshop artworks a long time ago myself and it's pretty time consuming. I mean we see it in a lot of job areas these days: machines taking over. But creativity? That's new to me though there was another AI image generator that generated some furore in the last months, don't know the name anymore.

ChatGPT is kind of a similar approach, what's going on that in a short time AIs popping up "everywhere"? Feels like an agenda to me.
 
Remains to be seen.. I think it will simplify some things, many things, and designers and artists will use their creativity in other ways that make use of this new tool. I think it will just change things, sort of like Photoshop did

HolySchmidt said:
would be interested in how the stock image looked like.
Click to expand...
i don't have it anymore but it was nothing crazy, anything you throw at it is fine, even blobs of color.
here's an old example I did to showcase that

1676586024482.png


HolySchmidt said:
ChatGPT is kind of a similar approach, what's going on that in a short time AIs popping up "everywhere"? Feels like an agenda to me.
Click to expand...
The agenda of making 💰 💰 💰
 
Okay, after playing around for hours and hours (this shit is addictive as hell!), I "created" something like this:

01982-2513527608.png


I've created tons of images with this style, it's absolutely phenomenal. I could do this for days and I don't know really why :D
But it definitely triggers my creativity. Thanks for creating this thread guys!
 
HolySchmidt said:
Okay, after playing around for hours and hours (this shit is addictive as hell!), I "created" something like this:

View attachment 213865

I've created tons of images with this style, it's absolutely phenomenal. I could do this for days and I don't know really why :D
But it definitely triggers my creativity. Thanks for creating this thread guys!
Click to expand...

Thats fantastic matey! Yup i've spent hours and days playing with SD haha, ill post a few albums up here shortly :p
 
snowkidy said:
Nice! What's your workflow to keep character consistency?
Click to expand...

Essentially creating embeddings or Lora's to keep the facial features in place and playing around with the Positive and Negative promptings to ensure i get a solid output, also i tend to avoid hiRes whilst creating and just bang out over 100 renders before i sift through the images and choose the seeds that suit me.

Its pretty much hit or miss sort of scenario, but i tend to get good results 9/10 of the time.
 
I've played around a lot more and read some stuff about prompts and searched a bit on the net, finding the right prompts and so on.

00166-284419525.png

I am still baffled at how crazy this AI art-stuff is. To be fair: in the end it's just a bunch of other artists works mixed together to create something new. You could call it inspiration, but I don't since it's not made by a human so to say. I know, that's kind of a phylosophical debate. But it's just TOO easy: with the right prompts you can create a TON of pictures like this, you're only limited by how fast your GPU can calculate.

I am really split about this. I think it's really amazing that I, as an art-noob, can create tons of wonderful pictures. But what about all the real artists? The ones who created REAL art? I mean who would pay for something that is free now? That only takes seconds to be done and is highly adjustable to every individual aspect you could think of?

I think this is the death of every digital artist. Really. This will kill a whole industry in my opinion.

@SinfulSeductionsVR
If that is really your wife, the head I mean, I would take this offline immediately. Just a good advice, but I don't think she gave her permission. Your choice at the end, but you should know that the internet does not forget.
 
HolySchmidt said:
To be fair: in the end it's just a bunch of other artists works mixed together to create something new.
Click to expand...
Some would argue with that I think. In a way what we as humans are doing is also mixing up together things we've seen and we try to 'order chaos' in a way that makes sense. AI lacks of course the subjective factor (emotion, message, etc.) but it does stand its ground imo in terms of what might pass as artistic creativity.


HolySchmidt said:
I am really split about this. I think it's really amazing that I, as an art-noob, can create tons of wonderful pictures. But what about all the real artists?
Click to expand...
Imagine the art artists are gonna create with tools like this though.. My bet is that it's just gonna shift things in a different gear. I think it's a bit similar with chatgpt and coding and writing. It can do some basic stuff for noobs of those fields but likely not anything great. While for people that are familiar with those fields it can help streamline their workflows and get more quality done quicker, let them have more fun and be more creative.

In some ways maybe it might elevate artists and increase the gap between noob work & quality work
 
SPQR said:
Some would argue with that I think. In a way what we as humans are doing is also mixing up together things we've seen and we try to 'order chaos' in a way that makes sense. AI lacks of course the subjective factor (emotion, message, etc.) but it does stand its ground imo in terms of what might pass as artistic creativity.
Click to expand...
You are right, but wrong as well. The point is: if a human wants to create art, it takes time, depending on how complex the artwork will be. If I create "art" with this kind of programs, it's always just some mouseclicks away, it takes minutes to create an artwork that would take hours if not days for a human. And we all know that time is money. That's why I think this will kill an industry.

SPQR said:
Imagine the art artists are gonna create with tools like this though.. My bet is that it's just gonna shift things in a different gear. I think it's a bit similar with chatgpt and coding and writing. It can do some basic stuff for noobs of those fields but likely not anything great. While for people that are familiar with those fields it can help streamline their workflows and get more quality done quicker, let them have more fun and be more creative.

In some ways maybe it might elevate artists and increase the gap between noob work & quality work
Click to expand...
I hope you are right, but one more point: with AI programs you are using the creativity of others for your own "work". I mean you literally suck it all in to create something new. The work of years probably is packed in some prompts so to say. You know what I mean? You could be inspired by all the best and greatest artist in one category, with all their work live in front of you. But mixing all their styles together in a new artwork would take days again. With Stable Diffusion I just mention their names and hit the "GENERATE" button. Done.

This is a totally different level of creating art. It's not a new tool, it's a new dimension. Imagine this as a 3D printer: you could spit out Picassos all day long, what is an original worth then? To be honest: all the artists and people whose photos are used or whatever these programs are using, should be payed everytime their pictures have been used for a new picture. That would be fair wouldn't it? A license fee if it's not marked as free to use.

I know their were law processes and that it's nearly impossible to say which pictures were used. But that's a problem isn't it? The question is: what is the value of art? Cause right now, with theses AI programs, digital art is worth nothing.
 
Last edited:
Generative art drastically narrows the gap between imagination and reality, but it doesn’t replace imagination.

What I've seen from this huge wave of AI art is the same homogenous crap. Even with powerful technology like this and you give it to normies or non-artists, we still get the same cyberpunk / greg rutkowski / blob of arts by people resharing prompts and pass it as their own art. The images look aesthetically nice but there is no meaning, no story, no coherence of what makes an artwork. What makes art in my opinion so beautiful is knowing that a human has sacrificed a good chunk of his life to acquire the skill to create that composition. When a machine creates it, the magic for some reason dissappears and this is coming from someone that has spend over 100s of hours in Stable Diffusion / Dreambooth and finetuning my own custom models.

As beautiful as the images are, an AI artwork will always pertain a hollow feeling to be passed off as "art". The fact that my GPU can generate 1000s of these images within an hour just devalues it in my opinion. It just feels too much like a slot machine to be called art in my opinion. Generative art has many use cases and it will definitely replace your local furry graphic designer that does $20 commissions but I don't see it replace fine art or high art at all.


Anyway here are some images that just freshly came out of my graphics card. I use the Deliberate v2 and it's my favorite model at the moment. PNG info should be embedded within the images so you can always easily reload the generation instructions
 

Attachments

  • 08992-2344618423-arcadian landscape, flower field, Flora, gaia, (beautiful young woman), full ...png
    08992-2344618423-arcadian landscape, flower field, Flora, gaia, (beautiful young woman), full ...png
    5.2 MB · Views: 0
  • 09638-2200672342-((beautiful young woman)), flower field, full body, porno, glasses, silver pi...png
    09638-2200672342-((beautiful young woman)), flower field, full body, porno, glasses, silver pi...png
    5.7 MB · Views: 0
HolySchmidt said:
I've played around a lot more and read some stuff about prompts and searched a bit on the net, finding the right prompts and so on.

View attachment 216350

I am still baffled at how crazy this AI art-stuff is. To be fair: in the end it's just a bunch of other artists works mixed together to create something new. You could call it inspiration, but I don't since it's not made by a human so to say. I know, that's kind of a phylosophical debate. But it's just TOO easy: with the right prompts you can create a TON of pictures like this, you're only limited by how fast your GPU can calculate.

I am really split about this. I think it's really amazing that I, as an art-noob, can create tons of wonderful pictures. But what about all the real artists? The ones who created REAL art? I mean who would pay for something that is free now? That only takes seconds to be done and is highly adjustable to every individual aspect you could think of?

I think this is the death of every digital artist. Really. This will kill a whole industry in my opinion.

@SinfulSeductionsVR
If that is really your wife, the head I mean, I would take this offline immediately. Just a good advice, but I don't think she gave her permission. Your choice at the end, but you should know that the internet does not forget.
Click to expand...

Definitely my wife, definitely gave permission, need not worry, we've posted much worse on a different site :p I won't go into tooooo much detail about that, but she's making her mark on the net haha

About the Art Industry, indeed they should be worried, with what we can achieve, its madness!
 
Finally got around and installed ControlNET and played around. Absolutely blew my mind couple times throughout the day. The level of control you have is crazy. Stable Diffusion always felt like a slot machine now it introduced creative direction importing images from Blender, VAM, whatever.

Tip: Go to Settings > ControlNet > set Multi ControlNet to 3 (or higher)
Now you can stack multiple controlnet layers. Use OpenPose model to get the pose right. Use Depth model to map out the depth / shape of the input image. If you want to extract small details from your input image. Try and use scribble, normal, msld or combine them with multi controlnet. Each have their use depending on how much creative control or freedom you wanna give to the diffusion model.

For anyone that wants to save on storage space and use just a .pt file of ~600 MB for the controlnet models. Someone in the community extracted them from the big 5 GB ckpt and it works with like a charm https://civitai.com/models/9251/controlnet-pre-trained-models
 
Yeah I am like 4 weeks in with this stuff, its really amazing.
Spent most of the time with it with one goal in mind: recreate an adorable co-worker.

So I trained my own checkpoint and lora with pictures of her, after two weeks I was pretty satisfied with the results.
Now training a TI as well with her, which is pretty heavy on time consumption and GPU (I own a 4070 TI OC).

But this is so addicting, if I finish my 1st goal 100% (recreating this coworker), I start to fiddle with Controlnet. The possibilities are overwhelming at the beginning.

Todays photo session with my current progress, I am getting close:

Saint00062-401207844-hq-scale-4_00x.pngSaint00074-285705711-hq-scale-4_00x.pngSaint00044-827214917-hq-scale-4_00x.pngSaint00060-1669097827-hq-scale-4_00x.pngSaint00070-527339220-hq-scale-4_00x.png
 
Last weeks I spent 100% of my free time with Stable Diffusion! Unbelievable!

Ok, where is a plugin to export poses, depth of field and normal maps for ControlNET ? The native export from VAM can create the perfect input for ControlNET. I am confident, it will not be very complex plugin. Create the plugin, pleeeease!
 
You must log in or register to reply here.
Back
Top Bottom