Imagine I have 10 objects, each with a 1k texture map, and I merge all those actors, creating, say, a 1k texture.
Is it correct to assumed that the resolution / quality will (roughly) of each model will be 1/10th of the quality of the the original 10, each with a unique 1k texture?
I was reading this documentation and slightly overwhelmed:
Yes, you do loose resolution. But think of it more as a Scene Render result then a “merging” of textures. it’s pretty much the same thing in the end.
The point of merging is mostly to create impostors - stuff for the HLOD system.
Thanks for the reply.
My challenge is that I have made a model of a big scanned skeleton in VR, and I want each bone to look good up close (have baked high scan data to low poly normal etc). Currently I have a texture for each bone.
If I combine the 100 bones, it will look bad up close right?
I’m assuming I can’t have 100+ textures in the engine 
Well, unfortunately each bone is different if you are dealing with scanned data. Instancing won’t help here.
If you combine the meshes, yes, you will loose resolution. However how much resolution you loose might actually depend on how the merging occurs.
If you want to do it yourself, in a 3d program, you will get better reuslts.
the anutomated system doesnt usually know that multiple bones can fit next to each other because the UV isnt full.
I would manually merge parts until the UVs are full. This wont necessarily give you a single item, it really depends on the scan data…
Also, as far as quality upclose. Maybe you could levarare Virtual Texturing an that extended mapping here…
Aha!
This is exactly what I was looking for.
I can optimize the UV’s in Blender prior.
Haven’t heard of virtual texturing so will look it up 