I found this video from Daniel, and thought I’d share it with you all.
Although I don’t understand any of the dialogue, I can more or less deduce what’s happening: A computer runs an FFT on an audio file to pull out what frequencies are being tapped at any given moment. It then relays that information to the piano, where said frequencies (or close estimations therein, since the piano doesn’t represent the whole spectrum) are struck to create as close to the original waveform’s sound as possible.
Practical implications have yet to be seen, but it’s definitely a fascinating project. If we could find an acoustic instrument, likely not 12-tone equal temperament as the piano is, that reflected a broader range of frequencies, we could get closer to replicating the original audio.
I think that might defeat the point though, as the missed frequencies are the very thing that give the “Talking Piano” its unique, albeit creepy sound.
Tweet