Okay done for the night I think! Got a rough shader set up to blend between multiple sets of baked textures, means I can add effects and stuff later on too.

Random dev doodle today… trying to get a lil python script set up that will generate a spatial photo from a 2D image. They work REALLY well in headset so far… So far I’ve got MiDaS generating the depth map, now trying to figure out the best way to write a displacement map in python so I don’t need to do anything in after effects.

First attempt did not work but looks cool 😂

Uhhhh nope

Nope but getting closer! Gotta figure out how to interpolate between the pixels somehow… right now it’s just shifting each pixel n-points left or right:

It is insane how well this works 😂

Well it works! Definitely works on some things better than others. Surpised at how well it works with things like drawings and 2D art.

I just put it up on gh here https://github.com/scooterrr/spatial_video_converter/ if anyone has access to a headset and wants to try it out:

I definitely need a more accurate method of aligning two phones... gonna try 3D print something. The disparity is a bit too high and it takes a sec for your eyes to settle on it ... I don't really have a way to monitor it until I capture, something else to figure out I guess!

It's pretty wild when it works though! The effect is really fun, everything just looks miniaturized.

This is a capture of the disparity on a standard Spatial Video from a single iPhone...

Then next to it, a capture from my janky two phone setup. I applied a bit of manual offset to each eye which made it a little more comfortable to look at, but still not quite right.

This feels insane!

Fun little Sunday project… been sick of lugging an enormous wetsuit dryer around to cons, it takes up a ton of space… so I made a something a little more portable out of a baby stroller fan and a hanger.

Beginnings of a lil light weight vtuber engine. Using OSC being transmitted from an iPhone, captured by a lil python app that runs a panda3d renderer.

Blendshapes and everything come across, just haven’t hooked it all up yet. Trickiest part was converting ARKit face tracking coordinates to something usable in panda.

It’s working!

Trying to make an AD mask thing 👀