We present a scheme of learning condition transfer achieving pairwise optimal transport, from a generic dataset of from {(x, c)} where a sample corresponding to a specific c is unique