Maxime Mantovani 3/5 : An Improvisation
On December 21, Maxime Mantovani was in the studio with Henri-Charles Caget on percussion and Victor Auffray on flugabone, for a session of improvisations for tests of the device he is working on, which combines an interface and prepared sound synthesis models. A crucial stage of the process, in order to check the tools developed and the reality of their use. Maxime Mantovani lets us discover one of these improv sessions, which, by his own account, is quite unique, as the instructions were to produce peaceful music…
We discover the control interface in front of Maxime Mantovani, with its four strange disks, somewhere between the dial of an old telephone and a DJ turntable. This interface is not yet fully functional, it is supplemented with four screens to display the waveforms and the different settings of the patch.
“During this improvisation," explains the composer, "the interface lets me manage two MuBu Granulate synthesis modules (developed at IRCAM by the ISMM team), a CSound Looper (which allows me to create sound loops), the direct inputs of the musicians, and three sound synthesis generators using artificial intelligence.
These three generators can, at my discretion, use five pre-trained models from the same neural network algorithm (the RAVE model) on five different sound banks: darbouka, daf, voice, flugabone, and a final sound bank with all the exchanges between the NASA base in Houston and the Apollo 13 Mission.
Photo Victor Auffray at the flugabone in the Ircam studios
"I can use these sources, or even an electronic sound that I produce myself, to make the pre-trained models generate sound by morphological transfer1. I can also record the sound generated by the AI, and re-inject it into an AI - which allows me to generate sound without suffering too much from the latency and delays imposed by the heaviness of the model."
This improvisation is the last one of the day and the three men agree to do "something calm": "Our improvisations are often kind of edgy," acknowledges Maxime, "and this time we wanted to explore something more relaxed, textures that develop over time."
"From the beginning, I connected the percussions to the pre-trained model of voice synthesis. When Henri-Charles rustles his bundles of twigs, for example, all these sounds trigger whispers and breathing noises, with almost human articulations, even if we are well aware that this vocal double is a bit monstrous. I find it very poetic.”
“At the very end of the improvisation, I wanted to create a mise en abyme of the NASA sounds. So, I routed the flugabone to the pre-trained model on the ‘Apollo 13’ sound bank and injected the result into a sound loop created with my Looper, after having transposed it into the extreme low register. This loop went round and round, with an accumulation phenomenon, and I reinjected all this accumulated material, again into the pre-trained "Apollo 13" model.”
“This is how, starting from nothing, we manage to build a musical discourse together as we improvise.”
Session d'improvistaion dans les studios de l'ircam
Maxime Mantovani, Henri-charles Caget et Victor Auffray (décembre 2021)
