Next Generation Performance Rendering -- Exploiting Controllability

Keiji Hirata and Rumi Hiraga

We believe that the next-generation performance rendering system should be able to refine and improve a generated performance interactively, incrementally and locally through direct instructions in the natural language of a musician. In addition, the generated performance must reflect the musician's intention properly. For these purposes, we propose a new framework called two-stage performance rendering. The first stage translates a musician's instruction in natural language into the deviations of the onset time, duration and amplitude of structurally important notes and second stage spreads the deviations over surrounding notes. We demonstrate sample sessions using a prototype system that contains a gouping editor and a performance rendering engine.

In Proc. of ICMC 2000, pp.360-363, ICMA.