Comment by marssaxman

#embed is faster because it skips the parsing step. xxd generates source code tokens, which the parser must then decode back into bytes; embed just reads the bytes directly.