Description
Adaptive text-to-speech (TTS) system has a lot of interesting and useful applications, but most of the existing algorithms are designed for training and running the system in the cloud. This thesis proposes an adaptive TTS system designed for edge devices with a low computational cost based on generative flows. The system, which is only 7.2G MACs and 42x smaller than its baseline, has the potential to adapt and infer without exceeding the memory constraint and edge processor capacity. Despite its low-cost, the system can still adapt to a target speaker with the same similarity and no significant audio naturalness degradation as with baseline models.