This sort of thing has been bugging me for months now.
My current unwrap/bake workflow is to apply smoothing groups after done retopologizing/low-res model, unwrap by selecting smoothing group faces and planar map/quick peel them, then bake.
The results work fine (as seen in the images), but I'm aware that this yields more shells, which means more vertex count in engine.
So now I've seen many people model, and then when they unwrap - they unwrap freely and place seams where they would be hidden - then use TexTools to assign those shells their own SG's...
This is ideal - but i feel it isn't accurate or correct and gives crap baking results. I find a problem with this because from what I see, if you unwrap a sharp angled mesh out into one shell - and now textools has given that thing one smoothing group (as seen in the image), wouldn't that look wrong smoothing wise, rather than six seperate shells?
Hope i've explained this clear enough, and apologies for the noob question.
Thanks!
Replies
As for sharp angles being grouped as one UV island, it's best to split them at 90 degree corners or bevel those edges. Beveling can't be done efficiently in some situations but it can allow you to keep more UV islands together with little to no increase in vertex count.
I have my smoothing groups assigned, THEN I uvw modifier, select faces by sg, and project/unfold and done... I don't assign a smoothing group while unwrapping, if that is what you are thinking.
Basically, I assign SG's prior, so it's easier for me to select faces, rather than manually select faces, unwrap, and TT script like you mentioned.