SinGlow: Singing Voice Synthesis with Glow: Help Virtual Singers More Human-like

Abstract

Singing voice synthesis (SVS) is a task using the computer to generate songs with lyrics. So far, researchers are focusing on tunning the pre-recorded sound pieces according to rigid rules. For example, in Vocaloid, one of the commercial SVS systems, there are 8 principal parameters modifiable by song creators. The system uses these parameters to synthesize sound pieces pre-recorded from professional voice actors. We notice a common difference between computer-generated songs and real singers' songs. This difference can be addressed to help the generated ones become more like the real-singer ones. In this paper, we propose SinGlow, as a solution to minimise this difference. SinGlow is one of the Normalizing Flow that directly uses the calculated Negative Log-Likelihood value to optimize the trainable parameters. This feature gives SinGlow the ability to perfectly encode inputs into feature vectors, which allows us to manipulate the feature space to minimize the difference we discussed before. To our best knowledge, we are the first to propose an application of Normalizing Flow in SVS fields. In our experiments, SinGlow shows the ability to make the input virtual-singer songs more human-like. The code of the SinGlow model is available at https://github.com/discover304/singlow.