Is AI the future of Indie RPGs?

[mondblut]

It's struggling to understand horses

So do 99% of humans who try to draw one.

She... truly is my son.

 

[hbm]

Previously you had to manually find photos and art belonging to other people, photobash them into your desired scene, do manual touch-ups, downscale them, palette squash them, and the upscale them using nearest-neighbor to get that kewl vintage indie pixel art point-and-click/chapter-screen look. And that's if you're a lazy plagiarist. Now you can just write a prompt, Stable Diffusion said photos/art together, generate 30 outputs in less than 10 minutes, take a piss, cherry pick your favorite (which will look better than the photo collage in your head did), and do the usual pixelizing stuff, substantially improving throughput. People (especially enterprising third-worlders) WILL exploit the hell out of this, like it or not.

 

[Dexter]

and do the usual pixelizing stuff

What I've seen people try to do more is de-pixelizing via img2img: https://sciprogramming.com/community/index.php?topic=2081.msg15543

Spoiler

I've been having a blast converting Sierra images to modern art pieces using AI.

See below some examples:

I have found that it works best for fantasy style scenes and landscapes. Luckily, a lot of sierra games feature these!

https://www.reddit.com/r/StardewVal...e_stardew_valley_inhabitants_went_to_a_photo/

Spoiler

Not exactly some of the most "fortuitous" art style choices here, but an interesting proof of concept for any potential "Remasters" or "Fan Remakes" nonetheless:

Spoiler

Btw. about creating tiling textures: https://github.com/hlky/stable-diffusion/pull/267

 

[Grauken]

and do the usual pixelizing stuff

What I've seen people try to do more is de-pixelizing via img2img: https://sciprogramming.com/community/index.php?topic=2081.msg15543

Spoiler

I've been having a blast converting Sierra images to modern art pieces using AI.

See below some examples:

I have found that it works best for fantasy style scenes and landscapes. Luckily, a lot of sierra games feature these!

Not exactly some of the most "fortuitous" art style choices here, but an interesting proof of concept for any potential "Remasters" or "Fan Remakes" nonetheless:

Spoiler

Btw. about creating tiling textures: https://github.com/hlky/stable-diffusion/pull/267

None of those AI-enhanced pictures looks remotely good

 

[Dexter]

None of those AI-enhanced pictures looks remotely good

 

[Zed]

I'm not wholly against the (partial) AI takeover.
There aren't many AA/AAA game devs that leverage actual artistry any more. The concept artists they hire are forced to draw the same shit over and over again.
And for indies -- RPG indies requiring character concept art, cultural concept art, etc etc, in particular -- this is an amazingly powerful tool.

 

[Mortmal]

and do the usual pixelizing stuff

What I've seen people try to do more is de-pixelizing via img2img: https://sciprogramming.com/community/index.php?topic=2081.msg15543

Spoiler

I've been having a blast converting Sierra images to modern art pieces using AI.

See below some examples:

I have found that it works best for fantasy style scenes and landscapes. Luckily, a lot of sierra games feature these!

Not exactly some of the most "fortuitous" art style choices here, but an interesting proof of concept for any potential "Remasters" or "Fan Remakes" nonetheless:

Spoiler

Btw. about creating tiling textures: https://github.com/hlky/stable-diffusion/pull/267

None of those AI-enhanced pictures looks remotely good

Seems the AI (midjourney) favor some people more, beside one eye it's pretty good:

 

[RaggleFraggle]

I’m getting subtle Omega Mart vibes whenever I zoom on the details.

 

[Sarathiour]

The armor look like shit.

 

[Viata]

It seems like I'm going from "old good, new bad" to "human-made good, AI-made bad".

 

[The_Sloth_Sleeps]

Previously you had to manually find photos and art belonging to other people, photobash them into your desired scene, do manual touch-ups, downscale them, palette squash them, and the upscale them using nearest-neighbor to get that kewl vintage indie pixel art point-and-click/chapter-screen look. And that's if you're a lazy plagiarist. Now you can just write a prompt, Stable Diffusion said photos/art together, generate 30 outputs in less than 10 minutes, take a piss, cherry pick your favorite (which will look better than the photo collage in your head did), and do the usual pixelizing stuff, substantially improving throughput. People (especially enterprising third-worlders) WILL exploit the hell out of this, like it or not.

I agree that I can see people using it to create plenty more shovelware. Still haven't seen any use for these generators that I can use in my game.
If I was really pressed I suppose this could generate character portraits (although theres been that capability for many years). Perhaps some strange looking loading screens as well...but probably not even that.

 

[CanadianCorndog]

I plan to use it extensively for texture generation and level filler, making NPC clothing and details, shaping terrain based on rules outside of main areas, auto UV mapping, etc.
Another thing I want to use it for is like an art director assistant. Correcting colour palettes that are slightly off, for example.
Anything that has rules or systems, perfect for AI.
So, it's really not too different from existing uses of procedural generation, just more advanced.
I will still make my own art and create my own levels etc. but it will be more like templates. Make the base textures, let AI fill in the varieties, UV map the assets, add details, etc.
I think it will also push me to be more bold in my art and design. There will be so much fantasy and sci fi and post apocalyptic AI shovelware in the future that if you start from a generic idea you will be completely boring and irrelevant. People will just know.

 

[Mortmal]

The armor look like shit.

Knees too sharp i guess ?

 

[agentorange]

I'm not wholly against the (partial) AI takeover.
There aren't many AA/AAA game devs that leverage actual artistry any more. The concept artists they hire are forced to draw the same shit over and over again.
And for indies -- RPG indies requiring character concept art, cultural concept art, etc etc, in particular -- this is an amazingly powerful tool.

First part is mostly my sentiment. But more so I am against it on principle but I don't think it changes much from the current trends. Concept art and all the visual designs in games has been increasingly automated over the years, which I've complained about for nearly a decade now and think is one of the main reasons why all modern games look so ugly and retarded, and most big budget games now days outsource stuff like concept art to third party companies, or asset creation to third world Indians, etc. This has even been happening with indies now, with people outsourcing art to things like Fiverr rather than having a dedicated visual designer. The AI (or as someone else put it better, procedural art generation) will just make this process more and more efficient and assembly line like, which the average NPC seems to be fine with since it means more things to consume at a faster rate. It's just a tool that mostly benefits big companies that want to produce more crap more quickly.

I disagree with the idea that this is somehow going to be a great boon for indie development, though, since part of the whole appeal of indies was that they represented something small scale, more personal and unique in contrast to corporate games. Now indie devs can also have access to the equivalent of 500 outsourced Indians at the push of a button. I think it's kind of the same thing as the effect that all the advancements in stuff like procedural generation had on indies, where instead of suddenly seeing a bunch of amazing large scale, well thought out RPGs with "infinite unique content", you instead got 6,000,000 Procedural Metroidvania Rogue-lite Souls-likes. Or the effect that things like easy to use engines and asset stores have had on the amount of shovelware. The only indie games in the past decade that were worth playing were the ones that still seemed to be made a by small, closely knit teams of dedicated people, that were willing to put a lot of human effort into making something they were passionate about.

It seems to be a trend that whenever a new technology like this appears you get a bunch of people saying "wow now I can finally achieve my dream project," (look at things like digital cameras and CGI in film making) and then they produce some asset flip tier garbage, because the kind of person who wants to put in less and less effort, and wants to automate more and more aspects of their supposed "passion" was never gong to make something worth looking at anyway. .

 

[Lady Error]

Tried generating a few song cover pictures based on this one (with a reverse keyword generator):

This is what the AI came up with:

 

[Krice]

I have tons of 32 x 32 tiles to draw, I wonder can it pixel art.

 

[Davaris]

I'm not wholly against the (partial) AI takeover.
There aren't many AA/AAA game devs that leverage actual artistry any more. The concept artists they hire are forced to draw the same shit over and over again.
And for indies -- RPG indies requiring character concept art, cultural concept art, etc etc, in particular -- this is an amazingly powerful tool.

First part is mostly my sentiment. But more so I am against it on principle but I don't think it changes much from the current trends. Concept art and all the visual designs in games has been increasingly automated over the years, which I've complained about for nearly a decade now and think is one of the main reasons why all modern games look so ugly and retarded, and most big budget games now days outsource stuff like concept art to third party companies, or asset creation to third world Indians, etc. This has even been happening with indies now, with people outsourcing art to things like Fiverr rather than having a dedicated visual designer. The AI (or as someone else put it better, procedural art generation) will just make this process more and more efficient and assembly line like, which the average NPC seems to be fine with since it means more things to consume at a faster rate. It's just a tool that mostly benefits big companies that want to produce more crap more quickly.

I disagree with the idea that this is somehow going to be a great boon for indie development, though, since part of the whole appeal of indies was that they represented something small scale, more personal and unique in contrast to corporate games. Now indie devs can also have access to the equivalent of 500 outsourced Indians at the push of a button. I think it's kind of the same thing as the effect that all the advancements in stuff like procedural generation had on indies, where instead of suddenly seeing a bunch of amazing large scale, well thought out RPGs with "infinite unique content", you instead got 6,000,000 Procedural Metroidvania Rogue-lite Souls-likes. Or the effect that things like easy to use engines and asset stores have had on the amount of shovelware. The only indie games in the past decade that were worth playing were the ones that still seemed to be made a by small, closely knit teams of dedicated people, that were willing to put a lot of human effort into making something they were passionate about.

It seems to be a trend that whenever a new technology like this appears you get a bunch of people saying "wow now I can finally achieve my dream project," (look at things like digital cameras and CGI in film making) and then they produce some asset flip tier garbage, because the kind of person who wants to put in less and less effort, and wants to automate more and more aspects of their supposed "passion" was never gong to make something worth looking at anyway. .

I watched mental outlaw demo stable difusion, and he was typing in artists names who's work he wanted to copy. Ultimately its going to happen anyway, so if artists ever want to get paid for doing the hard work it takes to create their own styles, they will have to ask for laws to protect their work from AI copying. Its up to them if they want to do that.

I have tons of 32 x 32 tiles to draw, I wonder can it pixel art.

https://playerone-studio.com/gan-2d-tiles-generative-adversarial-network

 

[The_Sloth_Sleeps]

I have tons of 32 x 32 tiles to draw, I wonder can it pixel art.

This was also similar to my first thought (or isometric tiles).
Seems like the answer is no.
Still waiting for evidence that it can do decent tiled textures (also seems resounding no).
What can I say? Its pretty good at drawing deformed horses though .

 

[RobotSquirrel]

From an AI Voice acting perspective, it definitely is the future for indie devs as now its actually affordable and reasonable to do. In the past it wasn't. ReplicaStudio is the only one I'm familiar with, seems really good in that you have support for emotional context and there are accents as well. But essentially it is still just TTS with some fairly major improvements. I'm considering using it on assumption that a better offer doesn't show up. Its not like Obsidian used this workflow or anything btw granted theirs was much higher quality. To clarify as well I'm referring to a technology aimed at indie game dev thus the pricing point I'm aware better quality exists but it's too expensive.

 

[Non-Edgy Gamer]

It's garbage, but it's garbage that amazes you at what it gets right.

(Ignore the dog in the last one btw. Apparently that's just a placeholder image and I took the screenshot before it generated the samurai woman it was meant to.)

 

[Non-Edgy Gamer]

I have tons of 32 x 32 tiles to draw, I wonder can it pixel art.

This was also similar to my first thought (or isometric tiles).
Seems like the answer is no.
Still waiting for evidence that it can do decent tiled textures (also seems resounding no).
What can I say? Its pretty good at drawing deformed horses though .

It can draw what it thinks is a tiled texture, but it probably won't match up the way a real texture would unless you clop part of it out and only use that. Haven't gotten it to do isometric except as a more complex image.

 

[The_Sloth_Sleeps]

I have tons of 32 x 32 tiles to draw, I wonder can it pixel art.

This was also similar to my first thought (or isometric tiles).
Seems like the answer is no.
Still waiting for evidence that it can do decent tiled textures (also seems resounding no).
What can I say? Its pretty good at drawing deformed horses though .

It can draw what it thinks is a tiled texture, but it probably won't match up the way a real texture would unless you clop part of it out and only use that. Haven't gotten it to do isometric except as a more complex image.

Yah not really any good that.
Can you try say pixel art worn stone, repeating?
For extra points add medieval decorations to the edges (think UI).

 

[Non-Edgy Gamer]

Can you try say pixel art worn stone, repeating?

No idea if that works or not repetition-wise. If you want a UI around it, it's better to do that separately and then photoshop the two together or something.

Don't get me wrong, it can make some pretty interesting scenes from games that don't exist, but they aren't that precise.

 

[The_Sloth_Sleeps]

Can you try say pixel art worn stone, repeating?

No idea if that works or not repetition-wise. If you want a UI around it, it's better to do that separately and then photoshop the two together or something.

Don't get me wrong, it can make some pretty interesting scenes from games that don't exist, but they aren't that precise.

Yeah this doesn't really save any work, probably the opposite TBH.
There's some procedural texture tools already that can do tiling stuff, I'm pretty sure so that would be the way forward with that one.
I have tried making some borders for UI, isolated but honestly a google image search and tracing in Gimp would yield better results/less work.

It does do some interesting mashups but in terms of asset creation its not really practical.

 

[dbx]

You people greatly over estimate artist jobs. And game making.
Yes, machine learning in itself is not artificial intelligence, but are you so sure you need AI to make game art? or 3D game asset? or even entire games?

People believed that the recognition of object from images would require intelligence, yet a simple machine learning alorithm ended up beating humans at it.