Abstract
Text-to-speech synthesis systems are usually deployed in different environments such as cloud-edge-end devices, which require fast synthesis speed, small memory storage, and low computation cost. However, early neural TTS models usually leverage deep neural networks that are usually with large computation/memory/time costs. In this chapter, we introduce the technologies for model-efficient TTS. We categorize these technologies according to the methods they use as follows: (1) Parallel generation, which can increase the parallelism of the computation and improve the inference (or training) speed. (2) Lightweight modeling, which aims to develop lightweight and efficient models with small model sizes, low computation, and fast inference speed. (3) Efficient modeling with domain knowledge, which designs efficient models by leveraging the domain knowledge of speech.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
