[Header image: An AI trying to remake my pixel art map in a more stylish way. I made it for inspiration, and curiosity.]
It has been half a year since I started, but I haven't made that much progress. Mind you over this year I have been creating more and more beta designs (I think they're fun). This is sort of procrastination, but it is also a unique opportunity.
While I am using AI to generate them, I want to edit them with a human (or at least a better AI) to create a cohesive style. That human touch, whether it is by me or not, is what will add to the beauty of the game.
Although the scale of which I am able to create them, will allow me to differentiate from other games. Not to mention the real difference isn't how it looks, but how each creature plays (balancing may become an issue).
Why am I doing it this way?
Tools can make or break the game. If you try to over rely on them, then you are doomed to fail. However if you are able to balance it out, then the results can be amazing.
As a multidisciplinary person I am probably one of the most capable person to make a game by myself. As being someone of many hats allows me to have the skills in many areas that I need. Although I'm certainly not an expert in all of them, and I also don't have the time (not really at least) to become one. Perhaps I can gain a couple new deep skills, but for the most part I should rely on AI or humans to pick up where I cannot.
Types of Generators
Long gone are the days where the only options were mediocre image generators. With RunwayML for example you can even have solid text to video generation.
While I think most of these things that the AI comes up with will need a polishing. Some things are so good on its own the polish would be just for consistency purposes really. When it comes to audio I actually think that there could be some really unique solutions. Like for the cries of each unique creature in game for example.
Text to Image
The obvious one here, but it has come a long way from just DALLE or Midjourney. With Nokemon, or even some other generators out there. I found around half a dozen that worked back in January.
There are some artifacts that need cut, and the symmetry is often off when it comes to these designs. However with some trims, polish, perhaps better shading in some cases, these base ideas could become phenomenal creatures.
I haven't had a lot of luck in the other types of images department. Such as with maps, or even locations (which it can do if prompted well). Those are more dynamic, and can take a lot of work to generate a good response. Still needing a human touch.
I'm really excited about this, as if I create a system of generating movesets based on type, location, and habitat. Then each creature could be added at rapid rates. Meaning that DLC's that actually add something to the game could be made.
Text to Sound
While still a bit newer to the scene. There are some cool music generators, and even pokemon cries generators too which can be repurposed.
I think that they made need to be cleaned up in a DAW, but how to go about mastering that is a challenge of itself.
Text to Video
Something that could be really exciting. I was having a hard time because I have this scene stuck in my head for part of the intro.
Sort of akin to the original pkmn intros with sprawling backgrounds.
I wanted a long ranging path on the left, with a line of trees on the left side, and a small fence to the other side. On the right you have large fields of wheat blowing in the wind.The camera swooping/zooming in towards the distance with birds flying off far away, and it being shot in golden hour.
That is a pretty descriptive prompt, which should make for a solid response. However what I got was only similar, and the one below you see was an experiment. I used the same prompt but added cartoon, and pokemon style. Purely out of curiousity. Almost has a ghibli vibe.
In the original release of this post on Ghost cms, I was able to embed the video and audio, but on Paragraph I am not. I'm sorry!
Here is how it works
When you use a tool it requires a solid text prompt. You can just say imagine this and this, and expect great results. Maybe one out of a thousand times it will come up with something you like. However these tools can get too expensive at that much generation.Not to mention the time wasted waiting. No, you must prompt it right.
It is a matter of how you Prompt
Most people have their own way of doing it, and that may become a thing down the line. People's own styling of prompting. Not to mention basing it off of artists of old like Da Vinci or Dali.
You may want to add the odd high res or 4k tag in there too, which could help with the result.
Adding a Human Element
By making the prompt require ingenuity and creativity, then it still leads to the conclusion that humans are important. Sure there are some jobs that will be replaced. That is how it is for ANY technology.
However new ones will take their place, and probably by a bigger magnitude.
More importantly is that it still takes a human prompting it effectively to come up with great results in less time.
I am an artist too
I don't want my work to be replaced. That is a terrible feeling, but I also write and create unique content. Not only is there not people out there talking about half the stuff I do on PolyInnovator.
There isn't even an AI that can write like me yet. Most article generators have a pattern that is detectible.
Ai won't be replacing you anytime soon.
More scale for small creators like me
By making use of the tools in front of me I am able to do more with less. I could create a game of such higher quality and scale, all because of some tools laying the groundwork.
While I'd rather have a team of people behind me, I just don't. Not to mention some of the ideas the AI came up with are so unique. While those are few and far in-between, they are getting better. Such as with Nokemon.
The rise of these tools can help indie developers rise up too. Accomplish more than what was thought before.
