Generative Spatial Synthesis of Sound and Music

ongoing

Machine learning is particularly well-suited to the processing of ill-defined problems, where complexity rules out direct solutions; it is therefore of paramount importance in the field of music creation, where compositional knowledge is far from explicit. Until now, the generation of sound spaces has often been neglected in the application of AI to music and sound, whereas 3D spatial audio is undergoing significant industrial development. The G3S research project, designed by and for musicians and sound creators, will take sound creation to a new level by enabling the generative synthesis of sound spaces in a context of man-machine interaction. This research will challenge the very notion of space in music and revolutionize practices: we will propose new ways of creating, modeling and analyzing sound by natively integrating its spatial dimension. G3S has four research objectives: 1) to design a unified operative representation of spatial synthesis 2) to propose a thesaurus and quantitative measures to describe the spatiality of sound 3) to generate sound spaces from machine learning 4) to design interfaces for exploring spatial synthesis through listening, semantic or functional description. We will form an international research network on the subject to consider what already exists and test our new proposals; we will develop open-source environments for each of the four objectives; and we will generate musical and sound creations using the project's results. There will be a wide range of uses, from artistic creation (installations, concerts) to video games and applications on phones or in virtual reality. Finally, the developments achieved will be integrated into the workflow of sound designers, and compatible with spatial audio standards.