Play:
A modernized game adaptation of Franz Kafka’s ‘The Trial’, set in a contemporary society governed by an oppressive AI bureaucracy. Your choices determine how you navigate, endure, or defy a system designed to break you.
The game was authored in the Kinexus game editor:
This is good. I liked it. For me, this is better than your other game, because it’s more the genre i like.
I like the graphics. They’re done well. I can’t see any glitches. When there are crowds, I see too much symmetry. This was true in the other game too. It’s an AI thing to repeat stuff. Maybe keep an eye on that.
Was AI able to create these scenes outright, or did you have to do a lot of post fixing?
A minor point. In some scenes, the picture didn’t load. But if i restart and try again, it worked. So this isn’t a missing asset problem. Maybe network error?
Here are some pictures from a dystopian future story i am working on (or was but it’s on hold currently).
You wake up in your apartment to find a dead cyborg in your bathroom. And you’ve lost your memory. You have no idea who she is, how she died or whether you’re involved.
I’m currently doing renders for my graphic art. Renders are a lot of work, but i do get precise control.
Wow, great shots. So many assets. I hope you didn’t model all of them. Are you using Unreal? I have a big project I worked on for years for Blender for generating into the timeline, but also intergrated in the 3d space, to do img2img or vid2vid processing. But currently it’s on hold, bc of Kinexus and broken py libs. GitHub - tin2tin/Pallaidium: PALLAIDIUM — a generative AI movie studio, seamlessly integrated into the Blender Video Editor (VSE), enabling end-to-end production from script to screen and back.
Gemini is good a character consistency, but poor at composing shots, typically it’s straight on a wall, with a centered character looking directly at the camera. Two shots are symmetrical. So, yes, that could be better, but the tradeoff is the character consistency.
Thanks! I mostly kitbash for my 3D. So I take 3D assets and make them into scenes. Or I take a base scene and change it. It uses the iray renderer. There’s a toon renderer as well, which is super-fast, but results don’t really look like proper drawings.
Currently I don’t use AI in my pipeline. But i believe it has a place somewhere in the production process. I’ve always thought that AI won’t give you the “whole picture” and must be used in conjunction with other methods.
For example, i have had some promising results with image->image using 3D as the input. With this idea, you can set up your scene composition and camera angle in 3D and have AI copy it in its output. Nevertheless, it’s still quite experimental, and I’m not seeing consistent results either.
Another AI limitation is that most pictures are quite small, say 1K or so. I like to make 3-4k renders, which do look very nice on large screens.
Here’s a couple of 4K renders, that i think have that “adventure tavern” look. Although i do envy the way AI manages to get that cinematic lighting feel. Which is tough to get with renders.
You’ll need to right click and open for full size.
Nano Banana (Gemini Flash) is quite good at altering existing images, if you ex. want to quickly try out a different light-concept - and then replicate it in your 3d app:
Prompt: “Change the style to a cinematic photorealistic shot of a dark and dirty bar in medieval times, with lightning to dimmed natural lights, smoke in the air light focus on the lady in the bar, night, add bar guests, moonlight though the windows to the left.”
A close-up of the woman - in style consistency - apparently it has picked up on the 3d game feel:
As for upscaling, there aren’t any good free ones out there. https://magnific.ai/ and topaz should be the best at upscales - but they’re commercial.
@jkj_yuio As this is going off-topic here, if you (or anyone else what to discuss these or other things further, you can also reach me on my Discord: tintwotin's server
Thanks for the picture transformations. I have joined your server.
