Does Dmitri's suggestion work more generally? e.g. suppose I have a finite complex of vector bundles on a singular variety. (I realize this should be a comment, but I don't have enough rep!) Actually, that direct sum seems unfortunate if the goal is to always get a line bundle....