What do you think of the explosion of A.I over the last few years?

Grey Swordsman

5 (100%)
I personally find it fascinating. I've used the image generators to make some interesting looking pictures just for fun. But other than that, I haven't exactly found them to be crazy necessary for anything that I'd want to use them for.

They really can't do faces when you give something like Stable Diffusion a prompt like "Dragon Ball Character holding a Lightsaber" and lightsabers spawn everywhere, it's pretty funny looking.

Have you guys found any neat or useful use cases for it? Or any certain stance you take on AI lately?

I think it's amazing.  It's already revolutionizing how creative works are made.  Stable Diffusion, ChatGPT, ElevenLabs...

A lot of people are understandably upset about jobs being replaced by it, but on the flipside, there's a benefit: creative types who don't have the budget to hire those people aren't going to be limited by their budgets as much anymore.  It's similar to how desktop PCs made "bedroom producers" possible: tons of musicians became successful producing their own work in their own homes, because they didn't need to hire professional producers, get exposure via being signed to labels, etc.

Check this out...


^Someone used ElevenLabs (a new AI-powered text-to-speech engine) to generate voice-acted lines for an NPC in WoW that didn't have voice acting.  And it sounds so real.  The most pragmatic criticism I've heard from people about generative voice acting is that it's not as emotive as the real thing.. but isn't it? The emotion in this example sounds good to me.

Now, on the topic of Stable Diffusion... my first impression was "well this is pretty neat" and that was about it. Went a few months without looking back into it, until just recently. Digging into the power of it, I see now there's so much you can do once you start experimenting with prompts, loading up different training models, using inpainting (I'm still trying to learn that), etc.

Speaking on personal experience here, I've been using Stable Diffusion to aid in producing game art this past week.

Here's an example of something Stable Diffusion helped me create for a game I'm working on. The portrait art, specifically:


What I did to create it was:

1. Use Stable Diffusion, specifically with the "DucHaitenClassicAnime" model.  The output looked like this after I used Photoshop to get rid of most of the background:


2. Piped that into a program called Pixelator and tweaked the settings just right.  I entered my game's color palette into Pixelator to ensure that the image conformed to the right colors and the right dimensions.  I also played around with the "Smoothness" setting to get a good balance of detail without getting too much "noise" (galaxies of pixels, for lack of a better term- bunch of random dots used for shading).  

3. Opened the result into a pixel art program (Aseprite) for cleanup.  Pixelator usually has a few color mistakes, especially with skin tones (it wants to give everyone orange skin half the time).  So I fix that up.  And then I fix up jaggy pixels, add a black outline (Shift+O in Aseprite), smooth out some of the "noise."

All in all, the whole process took me around 15 minutes.  While I dabble in pixel art, drawing something like this myself would be out of my ability, so the other option would be to hire an experienced artist for 2-3 hours of work.

I've already blown through the art budget I had for this game, and giving this random NPC a dialogue image would be super low-priority anyway.. so without Stable Diffusion, this character never would have had a dialogue image at all.  But because of Stable Diffusion, I can give this character, and pretty much every NPC in my game, a dialogue sprite. It's not much, but I feel like it adds a lot more polish to the game.

There are other things I've been using it for as well -- set pieces, like this statue and this rock:


Granted, I could do a little more cleanup on both of them to make them fit the style of the rest of the art better (they are quite a bit more detailed), but still. I'm happy enough with how they turned out to leave them like that for now, and maybe adjust later.. just focusing on higher-priority dev tasks right now.

Excuse all my rambling lol. This is something I'm excited about: up until now, I've been severely limited by budget because my game is so large. Even though I've spent over $10k on it so far, I've always wished I had more money to spend to get a wider variety of art.. more set pieces in tilesets, more dialogue images, etc. Now it's finally possible. That's what I mean by evening out the playing field: creatives who have dreams beyond their budgets can finally see those dreams realized.

Last edited by a moderator:
I read the whole thing, I don't consider that rambling at all haha. That blue haired anime grill looks pretty dope by the way. Also like how that angelic looking statue looks too, I think it fits in your game nicely.

I read the whole thing, I don't consider that rambling at all haha. That blue haired anime grill looks pretty dope by the way. Also like how that angelic looking statue looks too, I think it fits in your game nicely.
Thanks man, appreciate that.

Another cool thing we're going to see get really good in the coming months I think is the creation of AI-generated 3D models. To be able to type in something like "knight carrying a sword" and have that turn into a 3D model, with the kind of power Stable Diffusion does for 2D images.. it's gonna be huge for game development.

I've googled a bit to see what's out there already for that. The public stuff seems a little crude (but maybe usable), but there are a bunch of whitepapers published. So it's gonna be really interesting to see what happens in that space this year.

GPT-4 was released today I believe. I'll have to check it out sometime, although I don't exactly know what to use it for.

On sort of a side-note, my very first interacting with ChatGPT-3 was to try and recover the very first mod I ever made titled "Irish Black Clovers" for Hearts of Iron 4. It didn't work lmao it told me it couldn't recover mods and that it had no knowledge of stuff before 2021. Something like that. Kind of a shame, but I tend to delete stuff randomly so ultimately that's on me.

I also remembered that GPT-4 can do image prompts, hmmm...

I followed a guide on YouTube to create some pretty cool images:

First one...

[attachment=37817:000007.ad9c66e8.1779710494 (1).png]





This one...


And this...


And also...


Almost done...


And lastly my personal favorite:


If you look closely at some of their hands, they do look a bit off. But some of the images that these models can create look pretty dang good!


Damn, those turned out really well.

What kind of GPU are you using?  I'm on a GeForce GTX 1070 (8GB), goes kinda slow for running Stable Diffusion locally. So I wound up switching over to a hosted service called Mage. Pretty nice alternative if you don't have a beefy GPU, as it takes only a few seconds to generate an image, and it also keeps a record of everything you generate (& the prompts you used), which is nice when you want to create something kinda similar to something you've already created. But the downside is you're limited to whatever they provide (in terms of models, etc.), so you can't download custom models/loras/extensions/whatever. They do have a pretty huge selection of models to use, though.

ChatGPT I've been using more and more. Even in my day job: a coworker asked me if I knew how to write a database query that would do a very specific thing, and I didn't know, but ChatGPT sure knew, lol. Such a productivity booster. Working on C# scripts for my game, too - sometimes I'll paste an existing script of mine and ask it to tweak a couple of things (often faster for me than looking up APIs I haven't used in a while and have forgotten specifics on).

Tonight I also pasted in a whole bunch of paragraphs from my game's design document and asked ChatGPT if there were any plot holes. It came back giving me a list of a few things in the story that I might not have adequately explained enough and should make sure to describe in more detail so as to ensure it makes sense to the players.

I'm using an RTX 3090 24GB card. It can spit out 5 images in less than a minute, but trying to go beyond the resolution of 512x512 (or whatever the default is) really makes the images look strange. I don't have the skills to cleanup the images, so I just downloaded what I thought turned out okay-ish. The main model I used from the link in the video's description had two "M"'s in the model title, I forgot what it was called.

I ended up uninstalling the whole thing since I couldn't use it offline like the YouTuber said it could.

Also, just looked it up and I'm pretty sure this is the same model I used: https://civitai.com/models/7240/meinamix

Daaamn that's a beefy GPU. That anime model looks pretty cool. The output there looks just like it was drawn by a real anime artist.

Sometimes, this stuff is even better than "real" art... I ended up replacing one of the monsters in my game this week with a Stable Diffusion-generated image, because the version SD drew just looked a lot better than the version a real artist drew.

This tweet today is just about the funniest thing I've seen regarding AI.  The thread is about a guy teaching ChatGPT to program itself.  Then this other guy asks the Bing Chatbot what it thinks about it, and the Bing Chatbot basically starts saying "the guy admitted he made the whole thing up, he never got ChatGPT to do that" and starts quoting fake tweets from him that don't exist.. LOL. It's as if these chatbots are getting smart enough to try and talk shit to downplay each other's powers  :rofl:


Last edited by a moderator:
Yeah, the images I do like too but ~3 of them look cross-eyed now that I've looked at them more lol. The beautiful thing about AI too is that it's all non-copyright-able (totally a word) it was proven in court very recently that there's currently no issues with it. So for instance, you could theoretically create AI generated music for a YouTube channel and not recieve any strikes. It's something I'd love to get my hands on to replace or add some music to problematic Let's Plays or something similar. The only website I checked out wanted people to spend like a minimum of $30 a month for access. 

Also on the topic of AI Art vs Commissioned Art, I can't afford to drop hundreds of dollars for one image. That's just not feasible. So I'm very glad that there's a free option for people with me that lack the skill. I'd love to make my own art still, but Adobe CC is also an expensive subscription. A large majority of the time I'd be learning, I'd also be shelling out cash.

I think at this point the barrier of entry has been lowered drastically for everyone that wants images, music, entertainment, basically anything digitally and beyond. I heard that they will be able to have Scientific breakthroughs just from AI alone in the future.

Those tweets were hilarious btw haha.

I think at this point the barrier of entry has been lowered drastically for everyone that wants images, music, entertainment, basically anything digitally and beyond
100%. The biggest shift I see happening this decade is that it's going to empower creative entrepreneurs, without budgets, to be able to see their dreams realized... movies, music, art, games that never would have existed will exist because of the lower barrier to entry.

Did you hear about the universal music group situation? It's absolutely disgusting behavior to just claim that AI generated music is not to be allowed anywhere. I'm sure you've heard, but if not there's another development in the AI field lol.

honestly I like it. Working as a copywriter It is very easy for me to generate really good and useful articles. So I guess I am not that one who ai really helps

I believe that eventually, AI may replace many people in their jobs. Already, I'm surprised by how much easier it is to reach a large number of people simultaneously using a text messaging system like this one https://clerk.chat/industries/nonprofit/ . The efficiency and scalability it offers could revolutionize communication in various sectors, including nonprofits, and potentially lead to significant changes in the workforce landscape.
Last edited:
bread's done