As already mentioned, Judith Deschamps is a visual artist. When she became interested in the figure of Farinelli and the aria Quell'usignolo che innamorato that the castrato sang every night to the King of Spain for years to overcome his melancholy, she felt the need to surround herself with musicians. Among the first to work with her and still actively participating in the artistic research project at IRCAM is the composer and countertenor António Sá-Dantas.
António Sá-Dantas, how did you meet Judith Deschamps and how did you come to be part of her project?
I met Judith four years ago, and that was already focused on the figure of Farinelli and the aria Quell'usignolo che innamorato by Geminiano Giacomelli. The first avatar of this long-term project was in fact an installation involving a live performance by singers. Judith had issued a call, to which I responded, being a countertenor myself and fascinated by the character of Farinelli.
During our first meeting, we talked about what there was to sing, but also about the visual aspect of the score as it was a central element of the installation. Being a composer as well, we discussed how the score could be modified visually to reflect Farinelli's extreme range and technical virtuosity. For example, we added lines to the staves, omitted the keys, and played with a gradient of grey for the notes.
Performance with Ana Beard Fernandez & Antonio Breitenfeld Sa-Dantas, Royal College of Art, 2018 © Judith Deschamps
What attracted you to this project?
First of all, the figure of Farinelli, whom I find fascinating, both in terms of gender ambiguity and formidable musical abilities. But I was also seduced by Judith's beautiful and touching approach, and the emphasis she puts on the relationship of the castrato with the King of Spain. This musical ritual that became therapeutic, even cathartic for Philip V's "melancholy". Finally, the relationship to new technologies was the last thing I was enthusiastic about. We are here in a real "sweet spot", where music, voice, mental health and deep learning come together - a technology that is so often asked today to come and make up for our difficulties.
This collaboration quickly turned into a friendship, and became an opportunity for me to be nourished by a great artist. Over the years, after having sung for her, and after having acted as a consultant on the visual aspect of the score, I wanted to get more and more involved, by participating in the development of the project at IRCAM.
What do you mean by that?
When Judith began her research work at IRCAM, the score had to be reworked to bring out the specific abilities of each of the singers involved and their range. Their parts had to be adapted to this very particular exercise that the technological system developed at IRCAM demanded, but without ever losing the feeling of the baroque idiom.
My background as a countertenor and composer was invaluable in understanding what was at stake, in trying to give musical coherence to the project, but also in translating Judith's thoughts through the score, the instrumentation and the ornamentation.
The discovery of a score of Giacomelli's aria, sent to the Empress of Austria and annotated in Farinelli's hand, made my work even more exciting! Farinelli himself had transferred (in color) his own ornamentation on the score, mainly in the repetitions of the theme and in the cadences. It is a rare document because it was not the custom of the musicians of the time to notate their ornaments: as a matter of principle, one improvised them, one did not write them down. This document is also particularly interesting, because it allows us to see Farinelli's range at this period of his life (which covered about two octaves according to what he writes - with age, castrati lost their high notes).
This discovery led me to revise the aria to adapt it (with its ornamentation) for the singers involved in the project and my work consisted of expanding Farinelli's two-octave range to nearly three octaves (adding the high notes he could reach when he was younger), to make the best use of the range of all the singers.
Extract from Quell’usignolo che innamorato score
How did you approach this task knowing that its result would be used for deep learning of an artificial intelligence?
The algorithm developed by Frederik Bous and Axel Roebel does not, strictly speaking, "reconstruct" the aria as sung by Farinelli. It relies on all the recordings it is fed to transpose the pitch of the parts of the song that the alto voice, which serves as a basis for this recreation, cannot reach naturally. This is why it is necessary to be very precise in the rewriting.
It is all the more complex since the singers sing a capella, without an ensemble, but must still be fully engaged, with tempo, energy, etc. It is also important to take into account the constraints of the recording and, if need be, the edits that will have to be made on each of the singers' contributions. I then helped Judith with the artistic direction of these recordings. In particular those of the alto voice, which, augmented by the AI, serves as a guiding theme in the "recreation" of the aria.
How far along are you in this work?
After having successfully recreated the first version of the aria, we are working on a new extension of the project. Judith has planned to present this work again in the form of a sound installation, and the idea is that a program could recreate—freely, and continuously—new versions of the aria, no longer simply with a voice that would follow the score like a "normal" performer. The program would autonomously recompose the piece down to its very structure, not simply recomposing the ornamentation or a voice that would follow the score. The listener will be able to listen to a piece that is constantly different, and yet always the same.
This again refers to the nightly ritual of Farinelli and the King of Spain: Farinelli sang every night, so it was like an endless song, and it is unlikely that he sang exactly the same way every night. This is another exciting development imagined by Judith.
Interview by Jérémie Szpirglas