The Process


SEPARATING INSTRUMENT STEMS


How might we obtain a track for each instrument in an orchestra?
I downloaded a copy of Sylvie Guillem’s final dance of Maurice Ravel’s “Bolero” from YouTube (courtesy of the channel “T :”). 

The audio, of course, was a stereo mixdown, and not spatial in any way. To make it “spatial”, we would need individual stems , or isolated audio, for each instrument. I achieved this in two steps: 

  1. Automatic Rough Separation: I passed the audio through the online stem-separation tool Fadr, which gave a stem for the Snare Drum, Strings, Wind instruments and Audience Noise. These categories have similar frequency ranges and character, so the tool was capable of extracting them as audio files. 
Fadr.com 


  2. Manual Cleaning: I imported the stems into Logic Pro X, then listened to the song from start to finish multiple times, manually separating individual instruments into their own track. The WInd section as the heaviest to separate, as there were 8 different instruments within it, but luckily they were in succession. 



SPATIAL “SET” DESIGN


How might we arrange each instrument’ speaker for the screening?
As with all sound design, the spatial audio experience had to support the performance of the dancer on stage, and only once it has achieved that may it tell a story of its own.

In order to decide where each instrument would be placed in space, I studied what Sylvie Guillem was doing at the time of entry for each instrument.  was able to rent stereo monitor speakers from the RCA Resource Store, which means two speakers for each channel. I would be able to give each pair a ‘mono’ signal, which both speakers would play equally. 

Based on Guillem’s behaviour, I grouped and assigned a location to each of the instruments, which correlated to the speakers’ location on the floor. I also assigned a “width” to the sound of each instrument, depending on its role and whether it was playing in tight unison with others, which translated to how far apart each speaker pair would be.

I ran all the sound through a GigaPort USB audio interface with 8 outputs, This is less than the instruments in the orchestra, so I had to group some instruments to fit into the available inputs. To compensate for this, I gave some groups extra width (the longer lines in the diagram above) to give the illusion of spatial presence..