adamabr

Interesting, I’ll look into that. I guess I was assuming making linked duplicates creates instances meshes?

Example:

I have this bearing I want to try to optimize as much as possible.

The bearing geometry is radially symmetric, however the texture on the front cover is not.

I make a copy of the original and keep only one radial slice of the bearing. Then I instance this so I essentially recreate the whole bearing again.

Shaded mode shows the issues of seams and textures.

Adding the Data Transfer Modifier and enabling custom normals solves the normal issue.

Adding a Data Transfer Modifier and enabling UVs solves the texture issue (in Blender anyway).

So in Blender it looks like we have created a perfect copy this way. Let’s export to Verge 3D.

The normals are transferred properly but not the UVs. Adding a batch geometry puzzle brings down the render calls to the same as the original version.

A reasonable question would be if this is even relevant to do. Is it worth it? (Assuming the UVs work) Here is a comparison of file size.

Only the original exported:

GLTF: 11 KB
Bin: 40 KB
Total: 51 KB

Only the instanced version exported:

GLTF: 22 KB
Bin: 3 KB
Total: 25 KB
Savings: -26 KB (-51%)

So my take is that there are a lot of potential savings here if we can figure this out. What do you think?

• This reply was modified 1 week ago by adamabr.
• This reply was modified 1 week ago by adamabr.

I appreciate the response David! I’ll have to look into this more as I have more time but I am happy I got some input.

Interesting find!

But why does it work using the HTML overlay without clicking the fullscreen button when there is no “when clicked” puzzle?

I find this to be very odd. Does the “when clicked” puzzle affect the focus somehow?

And what workaround could I use?

Thanks for the rapid response kdv.

However, I’m not sure I agree. Try removing the “when clicked” puzzle and it will work both ways. I think there is something else going on here.

That is true, however it does make a big difference in my scene. The only performance problem so far is when it is switched for the first time. I’d rather take a longer loading scene and have that be smooth later.

Wonder if we have any control over this?

I am not afraid to delve into code and mess around.

Thank you for taking your time to look into this Pavel! Very interesting to see your methods.

I’ll share what I find from my experiments coding custom shaders.