Yes. I would start by brainstorming the setting and the idea and see if GPT can help with suggestions.
Something like this:
Take a low-effort, incremental approach: do something small as soon as possible, then improve it as we have time and interest. I think the fact that it’s just a mockup would help avoid the common “it can’t be released because it’s not ready” trap. Let’s just release a half-baked mockup and see what happens.
Brainstorm plot and environment ideas. The plot doesn’t have to be complex: it’s just two rooms. It’s possible to mimic a brainstorming session with GPT and see if it can provide interesting suggestions. I’ve done it for serious topics that had creative aspects, but I’ve never done it for fiction.
We choose a plot idea that we think will lead to cool environments generated by Midjourney.
Let’s stop there for now and see how it goes.
Yes, it’s likely that some manual work will be needed, but that can be quantified after seeing what has been accomplished by Midjourney.
Edit: I’m mentioning Midjourney, but it could be another similar technology. It doesn’t have to be decided right now.
I’m really stumped as to how the MI3a video art was made… I’ve been typing stuff into different AI art things for the past half hour and I’m getting nowhere near the look/feel.
It’s difficult to know what art style to ask for.
Even if it’s trained on MI2 art, the MI3a art is like that but more detailed somehow.
Yes, I wrote “screenshots”, but I meant “illustrations”, meaning that it doesn’t really matter how you make people beg for an actual game, as long as your criminal mind manages to create non-existent cool-looking games.
I had some fun with rundiffusion.com, which provides 30 minutes of stable diffusion for free. I couldn’t find out how to do pixel art, but the result is nice.
here are screenshots of my sessions. I tried several checkpoints. I still don’t understand how to use img2img; I’m not sure my input picture was actually used.
Yeah, on both of them I messed with the contrast/brightness… the second image I put contrast up and brightness down. I looked at some shots of MI2 just to judge it and try to get it similar.
On both of them I did the pixelation first before doing the contrast/brightness, I don’t know if that makes a difference.
It’s weird how the houses and even the ships are basically fine (and insofar as they’re not it falls within cartoon license, like the trees growing in one of 'em) but there’s some kind of horse without a head or body? A guy with three legs?
Question to @Paul and @seguso : do you feel that the amount of detail in the backgrounds of the anonymous video is similar to what can be achieved with Stable Diffusion or Leornardo.ai?
I’m trying to understand what technology was used for that video and my impression is that, whatever it was, it should be able to produce both high resolution images and detailed images.
