More by trial and error than any formal decision, I've found that I'm getting MUCH better results by stitching my 360s first to a spherical 2 by 1 representation, and only then convert to QTVR or cubic. Stitching directly to cubic sometimes works (it's also much faster), but I consistently get better quality by going to spherical first. Only drawback is sometimes having to work on the "North pole" of the cubic result, as pixels there are heavily distorted in the spherical representation.
Is this something other users are also finding?
Stephane
