Continuous-time speech synthesis model handles timing like a physical process

Original: Continuous-Time Acoustic Modelling with Neural Controlled Differential Equations

Source: arxiv.org ↗