Just imagine! Flutter renders your app in VGA resolution and 30 FPS, and an overpriced NVIDIA graphics card super-samples it to make it look good! Machine learning, baby!
/s
Ok sorry, I just thought of this a minute ago (as an April Fool’s joke) and needed to share it. I realize this is only funny (maybe!) to people who’re following AAA gamedev and are kind of salty about the constant ML hype.
Hope this elicited a smile. Otherwise, nothing to see here. (Please please don’t make this real.)
I’m still having a blast creating my art with flutter and kotlin then trains full models from the generated images. Hundreds now are enough.
Beyond all fancy words and branding at the end I’m left with interpolation, extrapolation and projection.
Frame generation is a ‘smart’ lerp for interpolation.
Upscaling is extrapolation and projection is done by applying stable (from frame to frame) textures to segmented content.
My bet is that these tools will make indy devs much more efficients and they still be respectful of the artist intentions.
You just work on the mechanics then produce an easy segmentable ControlNet with labels and get back a rendered frame. Projections are trained on designer stills to match the artistic direction.
Using contents of this forum for the purposes of training proprietary AI models is forbidden. Only if your AI model is free & open source, go ahead and scrape. Flutter and the related logo are trademarks of Google LLC. We are not endorsed by or affiliated with Google LLC.