Game dev

CPZ devblog 008


We’ve been working on portraits for quite some time.

As the development was going forward, we quickly realized that a small indie team would require a whole department or a dedicated artist to achieve a high-level of non-generated real-looking portraits, so some solution in-between that utilized the AI tech of some kind came front and center.

And it’s been real fun playing with different portrait approaches, generators and engines, but the solution ended up being many, not one. So, after almost all was said and done, and most portraits had a pass, we still couldn’t decide that this was good.

But we did realize one thing.

Handle AI with care!

And went back to pixelarting all the process (is that a verb? should be a verb…). Pixel art is quick. Stylish. Fits the game design better. Makes everything look and feel nice.

We can say we honestly looked into the hype, all while keeping in mind that our game is planned to be an artistic achievement of sorts, not an industry generated product.

While the brave new AI frontier has opened up a world of creativity, at the end of the day – “It just looks a bit like a public domain science distraction projectArt is still important, but we seem to have forgotten about it these days” – as Alex Djordjevic, our lead game designer retorts.

Alex continues: “First time I opened an AI portrait maker to test it out, I realized I was staring at Bradley Cooper, the actor. Well, more a person LIKE him, his long-lost twin brother or something – but not him exactly. But obviously him. The ML model used his photos for reference, and deepfaked him over a placeholder portrait. Ok, I thought, let’s play. If you’re an oldskool demoscene dude like some of us – you’re looking at everything as a challenge since you’ve been to a scene party competition or two. You can draw this shit in Aseprite or something just by deciding to draw it to look like anyone. So, I though – great references – but let’s just pixelart this.”

AI remains to be a really good thing to use the same way you would use a model IRL. Put the person in desired pose you want, and then draw/paint/pixelart it.

As it turns out the big philosophical question is about artist integrity. But if you have time for big philosophical discussion, and the outcome is painfully obvious (art is real > ai is a compound distortion of real) – just go and do what you like.

Of course, some will argue the morality of things. Or where’s the fun in that? There’s legions of stoners just getting high and putting prompts in all the dall-E’s, midjourneys etc. we’re sure. There’s also zero doubt that a lot of the AI tools will take you to a certain level and you should definitely give it a go – but none will provide the satisfaction, look and feel across the board of doing it all as your own art. I mean, even if it’s not great, it’s still yours. And you can draw your team mates in there, too. Fun for the whole crew!

In conclusion – we definitely see AI image processing as a tool that is here to stay, but we’ll always prefer taking that pen and drawing that sh*t. Keep it real, and make it beautiful!