A Novel Method for Vietnamese Voice Cloning

Voice cloning is a task of speech synthesis technology, which reproduces a person's voice while retaining their unique speech characteristics from text. Speech synthesis technology has developed rapidly in recent years. Some voice synthesis models have become popular and widely used. However, these models are primarily used for English. Meanwhile, research on Vietnamese speech synthesis remains limited. They only provide text-to-speech services with voices from specific individuals and cannot generate voices with unique user-specific sound characteristics. In this work, we present a method for Vietnamese voice cloning with the goal of replicating voices that retain the unique characteristics of the user's voice. To address this problem, we first build a dataset for the Vietnamese language. Then, we use this dataset to build a model for voice cloning. Finally, we use the built model with text as input and an audio segment of the target voice to generate synthesized speech that retains the unique characteristics of the input audio. We have conducted evaluations with the proposed solution, and the results indicate that the proposed method is promising for Vietnamese voice cloning.