AI generated art

[XcentY]

Since my visuals don't resemble the style of what you posted here, it probably doesn't make sense to show something I made. But you will find many creators here who do this style, and considering that they are able to create about 500 images per month and render them for their game, I think that I have not written something unrealistic (You can often find visually processed games like this on itch.io if you would like to see some).


I can also predict that when you release the game, it will be extremely successful and you will earn millions. But the reality can be completely different. So we'll see in 2025

500 images a month without the help of AI ?
I doubt...
Don't say they use tools like photoshop, After Effect, Blender, Unity, Unreal Engine and so on... It's full of AI helping tools...

 

[Deleted member 5954819]

500 images a month without the help of AI ?
I doubt...
Don't say they use tools like photoshop, After Effect, Blender and so on... It's full of AI helping tools...

I didn't mean Photoshop; I meant pure software like Blender or DAZ Studio. And maybe it sounds incredible, but I am writing to you about real numbers.

As a perfect example, I will mention a developer named DrPinkCake, the creator of the games Being a DIK. In the nine months it took to develop his last update, 5000 static renders and animations were created (according to the information provided via his Patreon). If you calculate it, it is approximately 550–560 images per month, and it must be taken into account that it is not only about creating the image but also clothes, light, environment, and similar things before you get to the result. In this time frame, he managed to do post-work (which I don't want to take into account since we don't deal with Photoshop here) and coding. And yes, by all accounts, he should be working alone.

Of course, it's definitely easier for him since he has over 15,000 Patreons, so he doesn't have to deal with the financial situation and problems of an ordinary person. But the fact remains that it is not unrealistic.

Edit: No, he doesn't do visuals in the style shown here by AI, but it's still some art and not exactly low quality, so I thought it was a good example.

 

[XcentY]

I didn't mean Photoshop; I meant pure software like Blender or DAZ Studio. And maybe it sounds incredible, but I am writing to you about real numbers.

As a perfect example, I will mention a developer named DrPinkCake, the creator of the games Being a DIK. In the nine months it took to develop his last update, 5000 static renders and animations were created (according to the information provided via his Patreon). If you calculate it, it is approximately 550–560 images per month, and it must be taken into account that it is not only about creating the image but also clothes, light, environment, and similar things before you get to the result. In this time frame, he managed to do post-work (which I don't want to take into account since we don't deal with Photoshop here) and coding. And yes, by all accounts, he should be working alone.

Of course, it's definitely easier for him since he has over 15,000 Patreons, so he doesn't have to deal with the financial situation and problems of an ordinary person. But the fact remains that it is not unrealistic.

I don't say it's unrealistic but I say that I doubt it's done without the help of AI.
Blender has a lot of computer-enhanced features and most of the work is done by blender...
Don't think I didn't use all these tools. I know them. I even made a game and programmed a Delaunay triangular in C#.
That's why I highly doubt what you're saying.
Also, here you're talking about a full project.
And certainly once you have all your 3d models, textures, and so on at hand. You'll certainly make scenes faster.
But you have to start the project. It's not like you come to blender and within 10 mins, you get me the first image of your game.
I can tell you by experience that first you'll take much more time to select your model, refine your model, and so on until you get what you want.
The same goes for texture selection, materials, lighting, shaders, and so on...
It's even longer if you start programming your own shaders.
But after that, once you've everything, then you can render new images much faster cause you already have all tools and ressources at hand.
Also when you make images for a RenPy game, you can render a lot of images where you only do a little change in the scene and render it again. Or you just move the camera to another angle and you render. Which means that on the same scene... You can render more than one image without changing "clothes, light, environnement, and similar things". It's not like generating a new scene at each iteration.
So you have to put your figures into perspective... If you give me a full 3d model with a template scene. I can also render something in 10 mins.

 

[Meaning Less]

programmed a Delaunay triangular in C#.

I never saw someone bragging about something that is used as an introduction exercise in computer science.

So you have to put your figures into perspective... If you give me a full 3d model with a template scene. I can also render something in 10 mins.

Well now you know why DAZ is so popular, people just download the models pose and make renders, probably faster than AI once you have the skills, because you don't have to spend so much time fighting against inconsistencies.

Which is again just another reason as to why AI isn't really replacing much.

 

[XcentY]

I never saw someone bragging about something that is used as an introduction exercise in computer science.

Well now you know why DAZ is so popular, people just download the models pose and make renders, probably faster than AI once you have the skills, because you don't have to spend so much time fighting against inconsistencies.

Which is again just another reason as to why AI isn't really replacing much.

First I didn't know that Delaunay Triangulator was an introduction exercice in computer science because I never studied computer science. I programmed everything myself when I had a need for it. So sorry I don't brag about it. I just say what I did. Nothing more. And if you want to show me your contempt, I will deal with it like I always do : ignoring
Second I sense that you just want to mock me since the beginning.
So I will from now on ignore your posts.

 

[Deleted member 5954819]

I don't say it's unrealistic but I say that I doubt it's done without the help of AI.

Ah, so until AI started to be used, how was it done?

From what I know in the case of Blender, it uses AI Denoiser, but it didn't always exist. If there are other functions that can be added to Blender that have something to do with AI, I'm not interested anymore; they're not coming from the original software but only as an addon or plugin.

DAZ Studio uses Genesis AI. These features allow you to create human figures with the help of machine learning, which can automatically fill in certain parts of the face or body according to the specified parameters. But it is also not an AI that will decide how your creation will look as a final result.

 

[XcentY]

I will stop here.
Since the beginning I see this topic is not about "AI generated art"
but rather a fight arena for some conservative 3D creators who want to make sure that everybody consider AI as non-art and recognize their 3D work by making other believes that AI is just a simple image library mixer...
Even that we adore them like godly figures in the 3D creation because they want recognition and success...
But for me, AI is generating art and often more appealing than the majority of the 3D scene creations we find in Renpy games where we often find the same 3d models in multiple games amongst different creators... No to say the content generated from ILLUSION softwares

 

[Meaning Less]

I just say what I did. Nothing more.

Just saying, It is something weird to throw out there since it is unrelated to the topic and sounded like you were using to prove a point somehow.

But for me, AI is generating art and often more appealing that the majority of the 3D scene creations we find in Renpy games.

Well that's very subjective, personally I'm extremely critical of DAZ renders and the lack of originality caused by them, yet somehow this recent "AI cg" invasion managed to rank even below 3dcg to me because now they not only removed originality but also removed consistency...

 

[XcentY]

Just saying, It is something weird to throw out there since it is unrelated to the topic and sounded like you were using to prove a point somehow.

Well that's very subjective, personally I'm extremely critical of DAZ renders and the lack of originality caused by them, yet somehow this recent "AI cg" invasion managed to rank even below 3dcg to me because now they not only removed originality but also removed consistency...

I think we could find common grounds... We're not far to think alike...
AI models consistency is coming through refined models like LORA models for Stable Diffusion.
And I even think that we'll see better results in the coming months.

Here is what the LORA models can do in Stable diffusion in term of consistency : (here 4 LORA models)


PS : Thanks to moderator

 

[XcentY]

I think I'm gonna try to do something very special with Renpy
Chat-GPT just inspired me.
Following Chat-GPT, it's possible to use python script in Renpy to generate dynamic images from Stable Diffusion
I want to try
It will take time as I'm not accustomed with Python.

Imagine a story in Renpy with new visuals each time you play

 

[Gabaw]

The timescale for AI is crunched so tremendously that by the time I'm finished writing this sentence, one single neural farm has already made more progress than a human could in a month. Add to that hundreds of farms, all at the same time, all receiving diverse input, all talking to eachother to refine, learn and also "teach" their counterparts, and you have what is effectively more brainpower than all humans on earth today could muster combined; all focused in on a specific series of tasks, which themselves will be created by the AI as it learns what its masters want.

It is tireless, restless, formless, soulless, amoral, and brutally efficient. It can destroy us simply by understanding us better than we do ourselves. No need for legions of increidbly mobile, heavily armed, nearly indestructible slaveshells that can also do their own processing and independent prioritization of tasks and targets. But there will be those too.

I'd say good luck to us, gentlemen, but I can only muster a premature goodbye. I see no way to avert what's coming.

 

[DuniX]

The timescale for AI is crunched so tremendously that by the time I'm finished writing this sentence, one single neural farm has already made more progress than a human could in a month. Add to that hundreds of farms, all at the same time, all receiving diverse input, all talking to eachother to refine, learn and also "teach" their counterparts, and you have what is effectively more brainpower than all humans on earth today could muster combined; all focused in on a specific series of tasks, which themselves will be created by the AI as it learns what its masters want.

It is tireless, restless, formless, soulless, amoral, and brutally efficient. It can destroy us simply by understanding us better than we do ourselves. No need for legions of increidbly mobile, heavily armed, nearly indestructible slaveshells that can also do their own processing and independent prioritization of tasks and targets. But there will be those too.

I'd say good luck to us, gentlemen, but I can only muster a premature goodbye. I see no way to avert what's coming.

I wouldn't say we have reached all that.
It's more like cramming a internet's worth of knowledge into a human brain and using that as a search engine.
It does not do any active thinking for themselves, it's more equivalent to the passive unconscious and dream like state.
Although it has shown some sparks of emergent behaviour that might lead to further development we are not there yet.

At the end of the day when GPT 7 rolls out it will already have all the possible Data and Knowledge it can scrounge up and be Super Trained as much as it can be trained.
That's when we are going to see the real limit of this technology and how we are going to prod and tweak it to generate even more interesting behavior. The future is going to be for the AI Brain Surgeons that Edit those brains as the value of Data, Knowledge and Computation goes to Zero as we will already have all the answers.

Instead of thinking of a Human Brain's Consciousness and Intelligence think more like Gigantic Building Size Brain that has the sum total knowledge of humanity with the Consciousness and Emotional Intelligence of a Friendly Dog.

In terms of God we will become Omniscient not Omnipotent. We are not there yet in terms of Omnipotence.

 

[kintarodev]

Been messing with it for months now, not for any particular reason (even less making a game) and all I can say is that any AI is far from having the skills of a pro artist. Much less a mind to create something really new. So far it's only good for replacing stock photo online stores, and even so... I would dare to say that I am starting to see some stagnation into the whole technology.

 

[RangerScout]

AI generated content will bes a game changer

 

[Deleted member 440241]

you have what is effectively more brainpower than all humans on earth today could muster combined

I think you've grossly underestimated the processing power needed to handle all the incoming stimuli the human brain receives. These AI's can't outperform the human brain (even ignoring that we have no idea how consciousness actually works) they're just hyperfocused on a singular task. They're also connected directly to the system displaying the image, where humans need to create the image piece by piece with their hands. I'm pretty sure most human artists could create faster and better than AI if their brains were connected to a monitor too.

 

[Meaning Less]

AI models consistency is coming through refined models like LORA models for Stable Diffusion.
And I even think that we'll see better results in the coming months.

I saw some of it but still not sure it will ever get there since stable diffusion is just too general for specific tasks like game development.

What I can see coming in the future though is a software that doesn't fully rely on AI but works with it to create more standardized models in different poses in a more precise and limited faction, without background images getting in the way of the main models, and with different layers so that clothing can be changed/replaced with more ease.

If we manage to get there then maybe we could have something to replace the effectiveness of the 3DCG stuff.

 

[XcentY]

I saw some of it but still not sure it will ever get there since stable diffusion is just too general for specific tasks like game development.

What I can see coming in the future though is a software that doesn't fully rely on AI but works with it to create more standardized models in different poses in a more precise and limited faction, without background images getting in the way of the main models, and with different layers so that clothing can be changed/replaced with more ease.

If we manage to get there then maybe we could have something to replace the effectiveness of the 3DCG stuff.

That's exactly what lora models is about. It places itself in the cross-validation and can be train on a model, on a pose, on a painting style, clothings or what ever you want...
I think generative AI is too young to show all its potential...
Sometimes, when I see the productions on stable diffusion, I think too much AI-art creators (or prompters if you find "AI-art" offensive) are using the same lora models which make the whole AI creation looks like something repetitive but this is also how it's going to improve because the more they test things in a particular model/environnement/style, ... the more they find ways to improve models. Some devs are currently having some testing on clothes... There is a huge Reddit post on the subject. We'll see

 

[Meaning Less]

That's exactly what lora models is about. It places itself in the cross-validation and can be train on a model, on a pose, on a painting style, clothings or what ever you want...

What I'm talking about is a tool that allows to you have more precise control to actually pose characters around and edit things in realtime like it can be done with daz.

Extra training is not enough because AI alone will never really get there unless it is paired with smarter and more precise non-AI algorithms.

 

[GGAdams]

Ai bros are so brain dead. There is no debate. It's not a discussion about your feelings about it. It's illegal that's it. Look what I found on reddit just NOW and tell me it benefits the art industry lol

And for people that still thinks it's not an issue, well guess what, it is and it's getting care of right now. So keep coping guys lol, you're the real heroes of the industry lol

You must be registered to see the links

 

[Deleted member 5954819]

Ai bros are so brain dead. There is no debate. It's not a discussion about your feelings about it. It's illegal that's it. Look what I found on reddit just NOW and tell me it benefits the art industry lol View attachment 2571140

And for people that still thinks it's not an issue, well guess what, it is and it's getting care of right now. So keep coping guys lol, you're the real heroes of the industry lol

You must be registered to see the links

They mainly still do not understand that they cannot use publicly available AI for commercial purposes. The chance of ending up in court with an extreme fine is several times higher than for using music without a license or stealing assets for your game. It looks very simple to them, but when they think that the creator of the AI is some random demented person who didn't pay attention to these things and doesn't have access to who and what the AI is using to defend itself against this, it must be a total fool.

But you know how they say... Who wants to go somewhere? Let's help him there. These pseudo-artists are laughable to me. Yesterday he could hardly turn on the PC, and today he is an artist alongside AI. So funny...or maybe sad.