Can AI be the Ultimate Sandbox? (Patreon)
Content
At this point I think we all agree that AI is here to stay. As many of you guys know, I am an actual developer and have been for many years. Initially I think a lot of people in my field were really scared about LLM's "taking away our jorbs" because they are pretty good at transpiling spoken (well, written actually) English rulesets and specs into code that sort of works.
To be honest, they've only gotten better during the last year or so, although they've sort of plateaued - which is fine I guess, everything has diminishing returns so why should LLM's be any different?
Anyway, this post isn't about using generative AI to make code, or even using it to play the game with you or write dynamic scenes (although that might be fun hey?) It's about using a much older technology - StableDiffusion, to generate image content dynamically.
Now, this isn't to say that QoS is all of a sudden going to be riddled with AI images. Actually, the free and exclusive versions will remain 2D and based (at this point, somewhat loosely) on DAD - the Dynamic Avatar Drawer by PerplexedPeach. As you'll see from my latest release, I'm putting a LOT of effort into increasing that 2D experience and making it better.
HOWEVER... This:
Anyone who has played recent versions of FreeCitie's "PregMod" version will probably recognize some of the above screenshot. I'm basically putting StableDiffusion integration into the game - starting with ComfyUI. I'm going to be putting A LOT of work (and hilariously, burning a lot of AI credits) to create a smooth and easy to use surface here for people to try to use these tools to create their own visualizations for characters and scenes.
That's right, I'm not going to be adding shitty AI images to the game, YOU ARE.
Basically what I'm hoping to do is provide a really cool playground for people to experiment and enjoy themselves, maybe one handed. Wink wink.
Here's just some samples that the game made on the fly recently in different modes. This is VERY early days, but it shows the adapter is reading the game data and trying to generate the correct tokens to the models.
"Drawn" style, starting clothes on ship.
Same scene but with more masculine stats.
A "sissy" in the tavern.
On the ship, but more "bimbo" body stats.
Reprocessing the same scene but with body sliders jacked up
An early attempt to do the cooking scene.
I tried photorealistic but to be honest, unless I can find a LoRA for clothing or costumes, this just looks weird to me lol.
Anyway, as you can see. There's alot going on here. I wish the image gen was faster, but I'm going to be doing my best once I get this working to implement a system that is constantly processing thing smartly behind the scenes and trying to deliver images when you need them.