Just use more multiplexers. you can have up to eight multiplexed outputs going to another multiplexer. You cycle through the first multiplexer and then the next and then the next. Basically you multiplex the out puts of your multiplexers. This gives you sixty four inputs with six control lines. Three control lines for the 'master' multiplexer (the output of which goes to your arduino) and three that are shared by the eight input multiplexers (the outputs of which go to the eight inputs of the master multiplexer) . I'm fairly sure it can run fast enough for musical proposes, if you are using switches you might also need to debounce them and this is were you are likely to see latency.