A Novel End-to-End Turkish Text-to-Speech (TTS) System via Deep Learning

Text-to-Speech (TTS) systems have made strides but creating natural-sounding human voices remains challenging. Existing methods rely on noncomprehensive models with only one-layer nonlinear transformations, which are less effective for processing complex data such as speech, images, and video. To overcome this, deep learning (DL)-based solutions been proposed TTS require a large amount of training data. Unfortunately, there is no available corpus Turkish TTS, unlike English, has ample resources. address our study focused developing speech synthesis system using DL approach. We obtained from male speaker Tacotron 2 + HiFi-GAN structure the system. Real users rated quality synthesized 4.49 Mean Opinion Score (MOS). Additionally, MOS-Listening Quality Objective evaluated objectively, obtaining score 4.32. The waveform inference time was determined by real-time factor, 1 s in 0.92 s. best knowledge, these findings represent first documented HiFi-GAN-based TTS.