Find out more

In the live performance Le Chant de l'Ombre, musician Benoit Carré (SKYGGE), a pioneer in musical creation with AI, and the digital art collective Cosa Mentale combine their worlds in a musical improvisation and experimentation with the new possibilities offered by AI and real-time voice cloning.

On stage, Benoit Carré is accompanied by Cosa Mentale's two visual performers, who act live on a screen. At the piano, he triggers voices with hybrid timbres, like an intelligent echo to the notes played. These songs are generated in real time thanks to an A.I. (SoMax, Ircam) that listens to what Benoit Carré is playing and selects songs by similarity from a corpus he has built.

Thanks to a real-time vocal cloning model, Benoit Carré's voice and the sound of the upright piano give rise to ghostly sonorities: hybridized vocal timbres of singers from the 30s, and the ethereal voice of a jazz singer from the 60s. These sounds are integrated into a set with ambient-electro accents. In this respect, the music echoes the idea of hauntology: the return of the past to the present, here through A.I., combining cloning and the generation of musical fragments directly from musical improvisation. Like an intelligent echo that emancipates itself from sound, musical and visual improvisation are interconnected.

The musician's movements are captured by infrared cameras, creating both real shadows and evolving forms. These evolving forms are generated and modified in real time by the Cosa Mentale collective. The visuals react to the sounds, gestures, and impulses of Benoit Carré, accompanying the musical improvisation in a fluid, ever-changing aesthetic. The interaction between sound, voice, and image creates a symbiosis that extends the sensory experience, offering a graphic, almost organic interpretation of music and voice.

The boundaries between organic voice, synthetic voice, presence, and shadow are blurred, opening up a whole new creative space in which randomness and improvisation combine with the inherent variations of A.I. systems, depending on their computing capacity and available memory.