we are working intensively with GS.
Our project will use Babylon for displaying.
Its not only GS but combined with a couple of other things (but thats somewhere else to tell).
Comparing the different views and displaing of the GS throughout the production pipeline we noticed quite a significant change when we move the GS into Babylon.
I am aware of GS formats, SH, ply, spz etc, some of the feature we also want/need to use.
Here I want to ask about the differences of the Babylon GS render, as it seems that there is a significant visual difference.
Attached an image which shows the splat displayed side by side in the viewers we use during our production.
Postshot (1), generation. Superplat editing Supersplat exported as webviewer (2), Babylon Sandbox (3) .spz file (exported from SS as ply and compressed with own tool), Babylon Sandbox (4) .ply file (exported from SS as .ply). … (The visual display in the project is identical to the Sandbox.)
@yamaciller, since this is GS, the cards should be unlit and we should not need to be concerned with tone mapping environment lighting. That said, we do have control over contrast, exposure, and tone mapping of the overall render with the Default Rendering Pipeline. So you can certainly put back any lost contrast or exposure from your original. It’s hard to say what last mile rendering profile is on Postshot and Playcanvas that may be contributing to the render you see, but we should be able to get closer with image processing our render. If you want to experiment in the sandbox, right-clicking the Rendering pipeline entry in the scene explorer will allow you to add a new Default Rendering Pipeline so you can experiment with settings.
Post processing is something we can and also do use for the final look.
We are actually using the possibilities to control the splats properites itself (contrast, brightness etc.)
But this does not give back the fidelity it seems to have lost from the point of Generation.
Yes, I am aware that it is not so easy to say, where the differences are, and in this case I only can/do compare visually (which is not very comprehensive).
Does BBL have the same alpha blending implementation (accuracy wise?)
Could there be a difference?
If possible, we will try to get a bit deeper into the GS format to see whats going on.
(BTW. I watched the Talk regarding GS Standardization, which was very interesting and informative! This goes a little bit into that direction :)) )
GS is just too interesting, we love it. Thank you for implementing it btw .
This was the main topic of my presentation at the Metaverse Standards forum.
Compression and quantization technics have been used to keep GS size under control.
Regarding colors, IIRC Compressed PLY from supersplat have more range than SPZ.
Color conversion between format can be problematic as well. I remember getting lower contrasts because a conversion was a bit problematic.
Without good references (standard), this will still be problematic.
This said, It’s always possible to try to improve rendering quality. Usually, when it’s possible, I compare the shaders to check I didn’t miss something. Using Renderdoc or Spector JS.
Another solution I’m just thinking is to generate a reference, export it and see how it behaves in other rendering engine.
It’s possible with Babylon to generate a complete GS, set splat color/position and export it. For example, generate splat with a color gradient and see if a screen capture gives the same colors.
This is a good opportunity to work on rendering quality for Babylon 9.
