Babylon vs glb comparison

[phaselock]

Just a general question (I’m looking to move some Seedborn’s assets into glb format for better perf). Studying the glb export results in the PG, it seems that animation groups target transform nodes of a skeleton.

My limits (tested with current babylon file format):
max character mesh: 25~30ish clones per scene
deformable bones per char: ~40
animations per char: ~20

With glb, I’m seeing per file per char export of an extra ~40 transform nodes. I believe these are for targeting by animation groups? If I need to clone 30 chars, each with its own skeleton (due to animation blending), does that mean an additional 30 x 40 = 1200 transform nodes + 30 x 20 clones for animationGroups?

I’m aware transform nodes don’t eat into rendering budget but the additional 1200+600 data do consume memory? even tho they are fundamentally identical?

Am I missing something? Cos I’m not seeing the gain from switching to glb (for animated characters) for scaled up deployments other than the file load size?

[phaselock]

On 2nd thoughts, is there a way to convert animationGroups via bjs to animation ranges on skeleton? This would drastically reduce memory overhead…hmm…

[labris]

Is it possible to share 2 files to compare and, probably, optimize GLB?

@phaselock

is there a way to convert animationGroups via bjs to animation ranges on skeleton?

I am not sure how to do it better by Babylon, but probably you may do it in Blender.

How do I get from Animation Groups to Animation Ranges

Hi @chrisrcampart, please do check out this guide and question post: Mixamo - Babylon.js Documentation (multiple animations) Problem Loading Meshes with Textures (includes a video of exporting 1 animation) I’m guessing that in order to use animation ranges, the animation keyframes in Blender should be placed one after the other in the same time track, rather than each animation having its own separate time track.

[phaselock]

@labris

Is it possible to share 2 files to compare and, probably, optimize GLB?

I have both types of files, same mesh, same set of animations. Both exported from blender.
a) babylon format: 8 MB - animation ranges
b) glb format: 2.7 MB at 1 frame sampling, at 2 frame sampling, file size drops to about 1.3ish.

glb is alr as optimized as I can get, think the issue is memory vs file load. glb will certainly load faster, but at scaled deployments, not as efficient as babylon? at least based on my read from the PG.