Why don't indie devs use AI-generated images as art?

[The_Sloth_Sleeps]

Pretty much this thread.

that's better than 99.9% of the shit arts degree holders produce though?

I don't like "art" either and I hate shelling out $50 p/h for game artwork, but that's the reality. And having seen these tools outputs its not obviously not changing anything anytime soon.

 

[rusty_shackleford]

And having seen these tools outputs its not obviously not changing anything anytime soon.

based on... what?
the fact that these barely existed a few months ago?

yeah, you're right, progress has been so slow.

 

[The_Sloth_Sleeps]

And having seen these tools outputs its not obviously not changing anything anytime soon.

based on... what?
the fact that these barely existed a few months ago?

yeah, you're right, progress has been so slow.

You are pretty optimistic. I'm much more skeptical of your projection into the future.
Linear progression isn't a common theme in technological advancement. We still haven't invented "the wheel" of AI.

 

[rusty_shackleford]

And having seen these tools outputs its not obviously not changing anything anytime soon.

based on... what?
the fact that these barely existed a few months ago?

yeah, you're right, progress has been so slow.

You are pretty optimistic. I'm much more skeptical of your projection into the future.
Linear progression isn't a common theme in technological advancement. We still haven't invented "the wheel" of AI.

You're right, exponential progression is much more common.

 

[Dexter]

And having seen these tools outputs its not obviously not changing anything anytime soon.

Yes it's really horrid, can you imagine, it doesn't get hands completely right yet with the first publicly available model?

Spoiler

 

[KateMicucci]

Let me guess, you think things like giving everyone a platform to speak is also good for globohomo?

I can't tell if you're honestly trying to make a point and failing, or just doing your regular disingenuous retard act.

If you don't understand how letting anyone create art rather than only the people in power with the money to pay artists is bad for globohomo, I don't know what to tell you.

something something "muh human factor" as he creates another cartoon pushing trans pedo agenda on kids

How the nerds who got filtered by Draw-a-Box and Loomis's Blooks think the future art economy will work:

1. Push button to generate goyslop AI art for free
2. People give you money for pushing the button

 

[rusty_shackleford]

Let me guess, you think things like giving everyone a platform to speak is also good for globohomo?

I can't tell if you're honestly trying to make a point and failing, or just doing your regular disingenuous retard act.

If you don't understand how letting anyone create art rather than only the people in power with the money to pay artists is bad for globohomo, I don't know what to tell you.

something something "muh human factor" as he creates another cartoon pushing trans pedo agenda on kids

How the nerds who got filtered by Draw-a-Box and Loomis's Blooks think the future art economy will work:

1. Push button to generate goyslop AI art for free
2. People give you money for pushing the button

you might want to have a doctor look at this case of butthurt, he'll prescribe you some copium

weird how artists claim they do it for the craft but can't stop bringing up money

 

[The_Sloth_Sleeps]

And having seen these tools outputs its not obviously not changing anything anytime soon.

Yes it's really horrid, can you imagine, it doesn't get hands completely right yet with the first publicly available model?

Spoiler

Your disease is that you are a technology fetishist and fantasist. This blindness caused by excessive masturbation to "AI technology" has rendered your brain incapable of seeing anything but what you wish to see. Hands or no hands, its already been outlined why this is not as fantastic as you think it is.

 

[rusty_shackleford]

I had SD draw you some butthurt cream, hope this helps

 

[Dexter]

Your disease is that you are a technology fetishist and fantasist. This blindness caused by excessive masturbation to "AI technology" has rendered your brain incapable of seeing anything but what you wish to see. Hands or no hands, its already been outlined why this is not as fantastic as you think it is.

I'm sorry that a self-driving car ran over your dog, but that doesn't change where this field already is for anybody with working eyes and where it's obviously going or will end up ~5-10 years from now. I hope you keep paying "$50 p/h for game artwork", nobody is trying to dissuade you from that. I hope you keep doing it and remain butthurt when we're at the Photoshop or Unreal/Unity stage of AI tools. It's both funny to see and it's your products that are going to have to compete with people using these tools in the future.

 

[Dexter]

https://dreamfusion3d.github.io/
https://dreamfusion3d.github.io/gallery.html

 

[J1M]

Imagine paying $100k for an arts degree and watching someone who tries at this for a few weeks now knows more about various art styles and composition than you do. All because being an artist isn't gated by being able to hold a pencil anymore.

I'm a fan of gatekeeping, especially skill-based gatekeeping, there's nothing good in lowering the standards. And you're wrong, most of the best results shown from any image generator were chosen by people with actual artistic background or at least some skill. Best proof for how wrong you are would be all these "comic books" made with AI generated images. It's clear that the authors think they made something special but it's also clear they don't even grasp the basics of sequential storytelling. At best they consider comic books to be a bunch of pretty pictures with repeating elements. Don't get me wrong, teaching yourself art without suffering (and paying for) the cancer of modern art education is great, but you still have to spend a considerable amount of time on it and there's no shortcut through AI generators.

Gatekeeping is important. But it's also how the comic industry killed itself by having the wrong gatekeepers.

I care about the quality of the output, not if someone used Lightwave or Maya or a text prompt to create it.

Artists have done a terrible job of improving their workflows and tools. On a typical mainstream game today the art budget is so high it employs 100 people and leads to degenerate situations like outsourcing or art costing more than the rest of the game combined.

Has quality improved? No.
Has innovation increased? No.

The teams are too big to do either of those things. These tools are exciting because they are disruptive and will reduce the size of art teams, which increases the chances of those other things.

 

[J1M]

And having seen these tools outputs its not obviously not changing anything anytime soon.

Yes it's really horrid, can you imagine, it doesn't get hands completely right yet with the first publicly available model?

Spoiler

A month ago AI was better than the average RPG portrait. Now it is better than the average oil painting. Should probably abandon work and go back to spending 20 hours of labor per image.

 

[J1M]

I had SD draw you some butthurt cream, hope this helps

Sad that it got better and isn't drawing that cool alien language anymore.

 

[rusty_shackleford]

we need a 'shitty artist gets mad at stablediffusion' megathread

 

[rusty_shackleford]

they're both AI generated

 

[Sigourn]

So I want to make coombat AI-generated images, where do I start?

 

[KateMicucci]

Imagine paying $100k for an arts degree and watching someone who tries at this for a few weeks now knows more about various art styles and composition than you do. All because being an artist isn't gated by being able to hold a pencil anymore.

I'm a fan of gatekeeping, especially skill-based gatekeeping, there's nothing good in lowering the standards. And you're wrong, most of the best results shown from any image generator were chosen by people with actual artistic background or at least some skill. Best proof for how wrong you are would be all these "comic books" made with AI generated images. It's clear that the authors think they made something special but it's also clear they don't even grasp the basics of sequential storytelling. At best they consider comic books to be a bunch of pretty pictures with repeating elements. Don't get me wrong, teaching yourself art without suffering (and paying for) the cancer of modern art education is great, but you still have to spend a considerable amount of time on it and there's no shortcut through AI generators.

Gatekeeping is important. But it's also how the comic industry killed itself by having the wrong gatekeepers.

I care about the quality of the output, not if someone used Lightwave or Maya or a text prompt to create it.

Artists have done a terrible job of improving their workflows and tools. On a typical mainstream game today the art budget is so high it employs 100 people and leads to degenerate situations like outsourcing or art costing more than the rest of the game combined.

Has quality improved? No.
Has innovation increased? No.

The teams are too big to do either of those things. These tools are exciting because they are disruptive and will reduce the size of art teams, which increases the chances of those other things.

Art budgets are high, and will remain high, for a reason. Good graphics and art direction sell games.

What exactly is your experience with artist workflows?

 

[rusty_shackleford]

art budgets have been increasingly sinking thanks to how easy it is to outsource the chimp labor to asia

because that's all artists are:
trained monkeys

 

[Strig]

Imagine paying $100k for an arts degree and watching someone who tries at this for a few weeks now knows more about various art styles and composition than you do. All because being an artist isn't gated by being able to hold a pencil anymore.

I'm a fan of gatekeeping, especially skill-based gatekeeping, there's nothing good in lowering the standards. And you're wrong, most of the best results shown from any image generator were chosen by people with actual artistic background or at least some skill. Best proof for how wrong you are would be all these "comic books" made with AI generated images. It's clear that the authors think they made something special but it's also clear they don't even grasp the basics of sequential storytelling. At best they consider comic books to be a bunch of pretty pictures with repeating elements. Don't get me wrong, teaching yourself art without suffering (and paying for) the cancer of modern art education is great, but you still have to spend a considerable amount of time on it and there's no shortcut through AI generators.

Gatekeeping is important. But it's also how the comic industry killed itself by having the wrong gatekeepers.

I care about the quality of the output, not if someone used Lightwave or Maya or a text prompt to create it.

Artists have done a terrible job of improving their workflows and tools. On a typical mainstream game today the art budget is so high it employs 100 people and leads to degenerate situations like outsourcing or art costing more than the rest of the game combined.

Has quality improved? No.
Has innovation increased? No.

The teams are too big to do either of those things. These tools are exciting because they are disruptive and will reduce the size of art teams, which increases the chances of those other things.

I agree with the first statement. When merit based gatekeeping makes way to basically any other type we get what we have now in many industries.

I cant't agree with this:

Artists have done a terrible job of improving their workflows and tools. On a typical mainstream game today the art budget is so high it employs 100 people and leads to degenerate situations like outsourcing or art costing more than the rest of the game combined.

The creative sector as a whole adapts extremely fast, the rate of course varies from creator to creator. But I wouldn't say the workflows and tools didn't change, and radically, even in the last couple of years. And I'm not that optimistic, I suspect that the new AI tools will entrench the worst tendencies in design and art and add a couple more.

 

[J1M]

How is it possible that you need less people to program a game engine that runs on multiple platforms and renders 100 frames per second than you need to populate that game with grass textures and NPCs?

It's because the artists as a profession have not been allocating resources correctly. Dragon Age 2 is a perfect example. They had to push that game out in only 18 months but the art team thought redoing all of the models they already had was how that time should be spent.

Instead of having fun sculpting characters they could have made dozens of cave and basement interiors so every cave in the game didn't use the same layout.

 

[J1M]

Imagine paying $100k for an arts degree and watching someone who tries at this for a few weeks now knows more about various art styles and composition than you do. All because being an artist isn't gated by being able to hold a pencil anymore.

I'm a fan of gatekeeping, especially skill-based gatekeeping, there's nothing good in lowering the standards. And you're wrong, most of the best results shown from any image generator were chosen by people with actual artistic background or at least some skill. Best proof for how wrong you are would be all these "comic books" made with AI generated images. It's clear that the authors think they made something special but it's also clear they don't even grasp the basics of sequential storytelling. At best they consider comic books to be a bunch of pretty pictures with repeating elements. Don't get me wrong, teaching yourself art without suffering (and paying for) the cancer of modern art education is great, but you still have to spend a considerable amount of time on it and there's no shortcut through AI generators.

Gatekeeping is important. But it's also how the comic industry killed itself by having the wrong gatekeepers.

I care about the quality of the output, not if someone used Lightwave or Maya or a text prompt to create it.

Artists have done a terrible job of improving their workflows and tools. On a typical mainstream game today the art budget is so high it employs 100 people and leads to degenerate situations like outsourcing or art costing more than the rest of the game combined.

Has quality improved? No.
Has innovation increased? No.

The teams are too big to do either of those things. These tools are exciting because they are disruptive and will reduce the size of art teams, which increases the chances of those other things.

Art budgets are high, and will remain high, for a reason. Good graphics and art direction sell games.

What exactly is your experience with artist workflows?

I programmed a total conversion mod with a group of 3D artist friends when I was in school.

I have experience working with Unreal, Unity, and Godot for hobby projects.

TLDR: everything from hand-entering vertices for a DirectX program to writing pixel shaders to using the modern tools of the day at a hobby level.

Please elaborate on your experience with art budgets.

 

[Bad Sector]

It's because the artists as a profession have not been allocating resources correctly. Dragon Age 2 is a perfect example. They had to push that game out in only 18 months but the art team thought redoing all of the models they already had was how that time should be spent.

In other words, Jeff Vogel was right :-P.

 

[rusty_shackleford]

How is it possible that you need less people to program a game engine that runs on multiple platforms and renders 100 frames per second than you need to populate that game with grass textures and NPCs?

because artists are lazy and do chimp work which is why they get paid 1/5th as much as someone writing the engine

 

[Dexter]

Can it generate animations for those 3d models?

Might be of interest: https://guytevet.github.io/mdm-page/
https://github.com/GuyTevet/motion-diffusion-model

Some other recent stuff: https://imagen.research.google/video/


These things have already gotten some simple implementation in the SD WEBUI's in the past few days:
Making AI Variants: https://arxiv.org/abs/2208.01626 https://www.youtube.com/watch?v=XW_nO2NMH_g
https://energy-based-model.github.i...-Generation-with-Composable-Diffusion-Models/ https://arxiv.org/pdf/2206.01714.pdf

Wow, daily optimizations that cut memory usage in half! Someone must be replacing Python code with C.

Might propagate down to stuff people use soon: https://ai.facebook.com/blog/gpu-inference-engine-nvidia-amd-open-source/

To address these industry challenges, Meta AI has developed and is open-sourcing AITemplate (AIT), a unified inference system with separate acceleration back ends for both AMD and NVIDIA GPU hardware. It delivers close to hardware-native Tensor Core (NVIDIA GPU) and Matrix Core (AMD GPU) performance on a variety of widely used AI models such as convolutional neural networks, transformers, and diffusers. With AIT, it is now possible to run performant inference on hardware from both GPU providers. We’ve used AIT to achieve performance improvements up to 12x on NVIDIA GPUs and 4x on AMD GPUs compared with eager mode within PyTorch.

AITemplate is a Python framework that transforms AI models into high-performance C++ GPU template code for accelerating inference. Our system is designed for speed and simplicity. There are two layers in AITemplate — a front-end layer, where we perform various graph transformations to optimize the graph, and a back-end layer, where we generate C++ kernel templates for the GPU target. In addition, AIT maintains a minimal dependency on external libraries. For example, the generated runtime library for inference is self-contained and only requires CUDA/ROCm runtime environments. (CUDA, NVIDIA’s Compute Unified Device Architecture, allows AI software to run efficiently on NVIDIA GPUs. ROCm is an open source software platform that does the same for AMD’s GPUs.)

Also Textual Inversion is down to 6GB VRAM and Dreambooth 8GB VRAM now (unfortunately the second doesn't work on Windows yet), although with higher Normal RAM cost (24GB): https://github.com/Ttl/diffusers/tree/dreambooth_deepspeed

https://dreamfusion3d.github.io/
https://dreamfusion3d.github.io/gallery.html

You can now also install and try the first implementation of this: https://github.com/ashawkey/stable-dreamfusion
It seems rough so far: