Results, Future Work, & Conclusion

Goals Achieved

Overall, I felt the system and performance were successful. It achieved goals I set out in various ways, detailed below.

Saxophone As Input Source

The first artistic goal was to create pieces with all material coming from the saxophone (i.e., nothing preprogrammed), and in which the performer is clearly engaged with the instrument, rather than looking at a screen or dials.

While work remains to be done on the tracking layer, this largely succeeded. I felt I was engaged with the performance very differently than in my previous pieces that used MIDI inputs, live-coding, and scoring, and the musical results of doing so were very different as well. I was also much more aware of myself and my body as a performer who the audience watches, compared to the situation when manipulating dials and the like.

To me, this was a very satisfying and artistically engaging way to create music, and is a direction I will certainly continue to explore in future works. I believe it was also a more engaging performance for the audience than is possible when they watch someone simply manipulating electronic devices.

Variety and Entropy

Another goal was to ensure the piece did not sound like a standard looper performance. When I hear looper performances, I mostly find the material overly repetitive, in that each layer repeats the exact same material for a long time until the performer replaces it. Additionally, the structure becomes very clearly strophic by definition.

I intended to avoid these issues by having the lines captured from the saxophone altered significantly before sequencing, including temporal changes (i.e., stretching); allowing them to layer short notes on top of each other, thus creating composite lines; and having them die out after some time without intervention by me (though I can intervene to speed it up). I feel this succeeded. On listening to the recording, I feel the listener can clearly hear that the synthesizer material is derived from the melodic and harmonic choices I make on the saxophone, but does not sound like a looper. In fact, in conversation with others after the performance, I heard that even educated listeners could not tell for sure that looping was happening.

Temporal Elasticity

Additionally, I experimented with the choice of having a steady pulse audible or not, providing me an audible road map for the loop cycles and beat spacing of the sequencers. For this performance, I ultimately chose to not hear the underlying main clock. The timing in sequenced lines is determined both my timing as a player and by the main clock tempo and loop point (4 bars at 70 bpm in this performance). In performance, the sequenced lines have clear temporal connections to what I have played, but as my playing is not (after the beginning) synchronized perfectly (or necessarily at all!) with the main clock, there is plenty of drift and rhythmic surprise. This succeeded in achieving a free improvised effect and a feeling of temporal elasticity.

Improvisational Surprise

Another goal was for the system to prompt me to improvise in response to transformations, pushing me to play material differently than I would in a solo performance or in an improvisation over a set background. This was also successful, with the transformations and temporal elasticity providing an effect not dissimilar to playing with other performers in a free improvisation context.

Compositional Structure

Returning to the problems of standard loopers, a deficiency (to me) in these tools is that they lead the improviser to a structure of simply adding layers gradually within a strophic form and then removing them similarly (what Steve Reich called the “tent structure”). By having note lifecycles and note probabilities, and having those under the control of the fader movements, I was able to introduce a more varied textural arc in the piece. I do not think I explored this as well as is possible, but it was successful at least in avoiding an obviously layered structure and in providing more rapid changes of texture.

Limitations & Future Work

I have numerous ideas and plans for improving the system and process of creating music with it.

Tracking Improvements

The principal deficiency in the tracker is that it does not consider a change of pitch within a legato line to be a new note. Fixing this will require significant new code, as there must be a dual heuristic for new note detection. At present, this forces me to play detached lines for what I want captured. This can be interesting, as I can have a capture pick up something that sounds quite different from what I have played, for example, triggering a three-note capture with a line of many more notes by choosing where to tongue. However, it does force me into playing more loudly and aggressively than I would like for the whole piece. Improving the tracker in this area will be a high priority.

Dynamic Range

I am able to achieve dynamic variety through the volume of sound and the density of note probability, however, I found that in performance I wanted to be able to bring the overall intensity down more quickly than is currently possible. At present, if I have added many takes rapidly and played aggressively, this material rings out for some time as the notes make their way through their lifecycles. While the number of notes that win the dice roll of playing or not can be changed with the foot controlller, they still go on for a long time.

I would like to be able to bring down volume, spectral content, and density more rapidly from gestural control, and to do so in a way that is reversible (i.e., not just killing off notes).

Gestural Controls

To achieve the above, a high priority for me will be developing additional input streams that can be manipulated while playing saxophone, without breaking the performer’s connection with the audience. Foot control of faders proved to be surprisingly effective, though it will take more practice to do without interfering with improvisation. I believe this was also engaging to watch for the audience as I do not need to turn away from the audience.

I believe a rich area of improvement will be mapping the physical orientation of the saxophone to input controls. The soprano saxophone is, compared to other saxophones, much easier to manipulate as a stick oriented in three dimensions, and can thus be used to conduct. Were I to change the microphone to a clip-on mic, and implement some physical tracking, I could influence the algorithms and mix automation effectively with multiple streams based on the position of the mouthpiece and bell and their orientation to each other.

Audio Controls

Another area that I have not explored in this version is having the note material captured by the system change algorithmic parameter or system behaviour based on the notes themselves. For example, pitch classes used within the last X captured notes could be used to determine key centers, and these used as input arguments to algorithms such as reflection and transposition. Similarly, measurements of the density of captured notes could be used to change texture.

Pulse & Temporal Interplay

While the temporally free approach in this performance worked, I would also like to explore works in which the relationship of what I play to an underlying pulse is more apparent. This could be achieved by having some indication to me of the underlying pulse, whether through a click track that is inaudible to the audience or through a visual indicator on stage (e.g., flashing light, small screen).

Another area of exploration will be to allow the pulse to change over the piece - at present the main clock is disguised, but is, in fact, static. A potentially fruitful area to explore will be allowing rubato pulses, with pulse changes coming from onset detection or gestural controls.

Additionally, the different track sequencers currently all use the same clock. Interesting effects could be explored by allowing their clocks to diverge.

Algorithmic Reworking of Audio

In the present version, the captured lines are transformed and rendered on synthesizer patches that deliberately sound nothing like a saxophone. While I don’t want the system to sound like a conventional looper, using the actual saxophone audio as the basis for electronically played material could be fruitful. This could include granulation of saxophone sounds, spectral transformations, or unusual stretching and delaying.

Additional Performers

Finally, I feel the system could be very rewarding to use with multiple players. This would also enable the performers to alter the algorithms more deliberately, perhaps variously assigning people to programmatic alterations versus improvising audio material. I also think composite captures that blend together input from multiple sources into single lines could be fruitful and provide surprises.

Additionaly, players could also help one explore different temporal styles. The system can send out clock sources to others, and could be effective combined with traditional sequencing or improvisation within a clear metric grid.

Conclusion

Overall, I feel the system developed for this project has been an artistic and technical success, providing me with a future direction that is both artistically exciting and will provide opportunities to highlight the advantages of working in Scheme for Max.

While far from perfect, the tracking layer is responsive enough for me to be able to add material quickly without worrying about the technical side. I found I was able to focus on the artistic output in a more concentrated, and dare I say, artistic, fashion than I have been able to previously. As an improvisor, I also felt engaged with, and excited by, the work more than I have for projects where I provide a composition, but do not perform or improvise.

I believe the continued development of the system and the addition of other performers will enable me to produce higher quality work with it and that the results will be interesting to performers as well as audiences.