Martn

Here you see a simple indi approach to 4DGS - just 10 cams - 8 in front - 1 above - 1 behind

​

Of course, behind falls apart, but the front, thanks to the used markers, got quite well matched ...

I had as well experimented on an very old 2021 recording where I ofourse didnt have any markers in place but the recoridng was done with 12 cams and the result is as well interesting it is not perfectly matched but the body is somehow closed. So another tiny step that even individual like me how are kind of in compare to the big offcial studio with big investment 4DGS are possible for XR storytelling ... 

I wrote again with AI code to handle the files and process the folders full of tiny 10.png image sets and experimented with sparse clouds and more dense clouds, I experimented with rotoscoping and different Gaussian creation tools ... It is a challenge to get "OK" looking 4DGS out of this tiny amount of cams but it is possible with dedication and a lot of tests where the best cam positions are ...

 

btw. I even experimeneted with SOG compression what is amazing the final folder full of the SOG stream is so tiny

that in the moment the format and an compatible player for unity or unreal with SOG-Stream function is available

another step for the future of spatial storytelling would be taken ...

 

anyway I still believe each technic Gaussian or Volumetric Video Mesh approach has its use cases and spatial storyteller or Volumetric Cinematographers need to get to know this new road ...

 

I understand my self still as an visual storyteller who has at the moment then focus on the Spatial future and mastering recording workflows to be ready for my future project or support others in there projects ....

 

I will speak at the upcomming Film Festival "Cameriage 2025" about it at stage ...

but more about this in another posting ...