This paper describes the winning systems developed by the BUT team for the two tracks of the First VoxSRC Speaker Recognition Challenge, we proposed r-vector in this paper. Update: I lanuched an open-source project wespeaker, where the implementation can be found