That's over 1 minute better than my best time, but it still looks like he could save nearly 1 second with faster turning on the first couple of turns! Those first few double-layer moves look a little slow. There's also room for a faster single without parity. He used the same PLL parity alg I use, I think, although I think it was just parity and no other PLL, which is probably better than no parity and most PLLs.