Dialing in optimal high (tri) to low (quad) retopo / projection vs ballparking it


Going from high poly tri model to low poly quad for animation.

Tuts keep saying to just get subdivided low poly count into ballpark of original (some say slightly higher).

The more I work with it, though, the more I wonder if that’s really an optimal workflow for a couple of reasons:

  1. Unless I’m missing something, projection remaps the subdivided VERTICES (as opposed to poly faces) to the positions in the original object, so it’s the point count that matters, not the poly count. The reason I bring this up is that if I have a high poly trianglulated model, then retopo in zremesh, the remesh will be quads, and the original will be tris. If I subdivide the remesh until it matches the poly count in the original, I will still be WAY short (nealy half missing) of matching the vertices in the original.

Correct me if I’m wrong, but doesn’t that mean it can’t project properly because there isn’t enough geo to project to all the target points? In other words, shouldn’t I be using the point count rather than poly count to dial in the projection numbers? On a related note, should the object be quadrangulated FIRST, and THEN retopo’d, using the quadrangulated high poly as the target rather than the original triangulated high poly?

  1. There’s a lot of general imprecision in the process as most describe it. Does it really not matter at all how much more subdivided geometry you have beyond that in the original object? Seems like there should be an optimal balance there somewhere if not 1:1 counts.

  2. If the low poly mesh has completely different topo flow (and possibly quads vs the original tris), isn’t it the case that even if I can map each vertex in the subdivided low poly 1:1 to a corresponding vertex in the high poly mesh, that when it MOVES, all hell breaks loose because the polys those vertices belong to don’t match in shape or flow between the two meshes?

If so, then wouldn’t it be a benefit to match as closely as possible the topo flow and geometry of the two meshes so that what we see after the projeciton matches the actual flexion of the low poly mesh as closely as possible?

If so, wouldn’t that look something more like retopo first at fully poly count, and THEN decimate down with no change in flow until low poly count is achieved, then divide back up in same amounts? Just makes sense to me that something along those lines would allow for better projection mapping, but:

It’s all new to me, so love to get some input.