Marking a year since it’s premiere, a score-follower video of a recent piece of mine, A Self-Portrait (Couperin) (2022-23) is now available on youtube, expertly performed by fabulous pianist, Adam Swayne.
Programme Note:
It’s not a self-portrait of me [MS] but it might have something of that ilk for a number of other parties. Derived from François Couperin’s B minor passacaglia (Pièces de Clavecin, Livre II, 8me ordre, 1717), the electronic component of A Self-Portrait is made from digital attempts to reconstruct the sound of the Couperin original in performance (a performance given by pianist Adam Swayne, for whom this piece was written) using audio grains from selections of Rolling Stone magazine’s ‘100 Most Influential Albums of All Time’. The live pianist finds ways to engage with the resulting mangle. An image of Couperin’s original drowning in the superabundance of music in our digital musical world emerged. Reaching and seething through some kind of viscous membrane. So maybe it’s a self-portrait of Couperin; or maybe one of Spotify.
Commentary:
Written for UK-based pianist Adam Swayne, A Self-Portrait (Couperin) is scored for solo live piano and fixed-media audio. In part, the piece is derived from François Couperin’s B minor passacaglia (Pièces de Clavecin, Livre II, 8me ordre, 1717) and the fixed-media component of A Self-Portrait is created using novel application of concatenative synthesis and automatic transcription. Concatenative synthesis follows a relatively simple workflow. A body of audio is analysed in software at a granular level according to preset parameters, usually variants on dominant pitch, pitch-to-noise ratio, spectral centroid (‘timbre’), and loudness. This pre-analysed audio forms the corpus for the synthesis. Next, a second (usually much shorter) piece of audio, the target, is selected. This too is analysed according to the same parametric values as the corpus. Software (in this case a patch in Max) is then tasked at finding the ‘closest match’ grain(s) in the corpus for each grain in the target, in effect, reconstructing the target using audio from the corpus.
How this is deployed compositionally makes sense in relation to my piece’s structure. The Couperin original is already rigidly organised as an introduction plus eight sections of relatively equal length. The introduction states a highly ornamented theme, then each of the following sections provide a unique episode of new material, followed by a restatement of the original theme. A Self-Portrait follows this structure, in fact (barring two moments of rupture), the piece simply comprises a single statement of the entire original in its original structural configuration.
A performance of the Couperin original was recorded by pianist Adam Swayne, and the recording split into the nine sections (the introduction plus each of the sections proper) of the original Couperin composition as described above. Each of these sectional recordings then became targets to be reconstructed from nine different corpuses (one corpus for each section, plus the introduction). Each corpus comprised complete audio from one of the top ten albums listed in Rolling Stone Magazines ‘500 Greatest Albums of All Time’ (Rolling Stone 2020), with the exception of the introduction, for which a corpus of similar length was derived, comprising all available recordings of the Couperin original listed on the streaming platform Spotify.
Using completed mastered audio recordings as a corpus presents a number of problems for the concatenative workflow. These, after all, are multi-layered recordings comprising (in most situations) simultaneous vocals, pianos, keys, and percussion (as well as additional FX and processing, following the given album’s aesthetic). Within such a context, the audible ‘meaning’ of the pitch, noise, and centroid variables available as part of the resynthesis process become much more questionable than in the case of more singular or monolithic sound sources. In most cases, the resultant resynthesis achieved in the making of my piece is a mangle of noise and glitch, to a greater or lesser extent far removed from the original target recordings.
But it should be noted that such results were very much part of the aesthetic game. The intent here was to create problems for the resynthesis process, appropriating the technique not in a mimetic but a transformative sense. On a conceptual level it interesting me that, in some way, that through the ‘eyes’ of the software these two audio passages are ‘equivalent’, despite the obvious auditory ‘distance’ their sonic character may provoke in a human listener.
The remainder of the live piano material was created from the resynthesized audio itself. The new audio was fed through mainstream commercial music transcription software. Again, the intent here was to create problems for the software’s algorithms. Such software is obviously primarily designed to take recordings of instrumental music, track the pitch and rhythmic content (usually through FFT protocols), and present these data as MIDI. Put simply: you’re supposed to put Beethoven in and get Beethoven out.
Obviously this presents problems when presenting the software with the dense noise-based material emergent from the concatenative resynthesis process described above. As such, the intent here mirrors that of the resynthesis itself, using methods of ‘transcription’ as further means of transformation. Again, it interested me conceptually that the transcribed result indicated somehow what the software algorithms ‘heard’. In performance, the resultant ‘transcriptions’ are performed alongside the resyntheses, synced via a click track.
A Self-Portrait also contains two moments of substantial departure from the original design of the Couperin. These occur at points 01:57-02:55 and 08:28-09:25. Labelled as ‘ruptures’ in my own sketches, these two passages are interjections or inserts into the flow of the Couperin, which otherwise proceeds around them unchanged. In the two ruptures, the live pianist remains tacet (the only two such moments in the piece). The audio is also slightly differently derived, although also concatenative in origin. Here, the target audio was ambient ‘offcut’ material taken from Swayne’s recording of the Couperin (where the mics were running between takes and simply capturing the creaks of a piano stool, footsteps, and perhaps distant, almost inaudible, conversation). This target was then reconstructed using the totality of all the other audio corpuses combined. The resultant resynthesis is momentarily sporadic-to-explosive but largely minimal (after all, the target audio can largely be considered ‘silent’). These sections are also the only areas of the piece where the resynthesized audio was further processed. Reverb and resonators were used to prolong and emphasise the tiny ‘bursts’ of sound in the raw resynthesized audio. The intent was one of momentary windows into ‘another world’ both sonically and referentially. A musicologist friend of mine described her experience of listening as suddenly and unsettling being cast from Merzbow to Vangelis.
When talking about this piece, I am often asked why I chose this particular piece of Couperin as a starting point. Part of the answer was strategic. The frequently recurring ritornello theme and sectional nature of Couperin’s work allowed for a sense of cross-comparison to be offered to the listener, as that theme reappears in different resynthesized forms. In addition, the highly ornamented nature of the original provides a challenging level of detail for the resynthesis process, encouraging glitches, falters and departures. But part of the answer is more circumstantial. Driving home from work, a performance of the piece was played on a BBC Radio 3 broadcast. Struck by the ornamental detail and harmonic patterns in the piece, the piece simply appealed to me. And, in essence, became ‘selected for me’ by the curators of the radio station, somehow akin to the curatorial selection of the ‘greatest albums of all time’.
A Self-Portrait Couperin was premiered by Adam Swayne at the Stoller Hall, Manchester (UK), on 20 August 2023.