From time to time I revisit my processes and self-made tools that create the XML and the tiles to keep them up-to-date. Since ever I have been converting the equirectangular image to cubes and then the cubes to tiles. For multiple resolutions of course. The following question has been asked a decade ago but I don't find a more recent one: Is this still the state of the art?
I don't mind converting to cubes first, but I'm wondering if using a sphere instead would be fine as well (plus split the sphere into tiles). Is there an advantage for one or the other way? Which one creates the best results, sphere or cubes?
Michael