Expediting TTS Synthesis with Adversarial Vocoding
This addresses speed issues in TTS pipelines for real-time applications, though it appears incremental as it builds on existing GAN and vocoding techniques.
The paper tackles the computational bottleneck in text-to-speech synthesis by using GANs to map spectrograms for faster vocoding, showing it outperforms naive methods and is hundreds of times faster than neural vocoders in user studies.
Recent approaches in text-to-speech (TTS) synthesis employ neural network strategies to vocode perceptually-informed spectrogram representations directly into listenable waveforms. Such vocoding procedures create a computational bottleneck in modern TTS pipelines. We propose an alternative approach which utilizes generative adversarial networks (GANs) to learn mappings from perceptually-informed spectrograms to simple magnitude spectrograms which can be heuristically vocoded. Through a user study, we show that our approach significantly outperforms naïve vocoding strategies while being hundreds of times faster than neural network vocoders used in state-of-the-art TTS systems. We also show that our method can be used to achieve state-of-the-art results in unsupervised synthesis of individual words of speech.