StyleTTS: A Style-Based Generative Model for Natural and Diverse Text-to-Speech Synthesis.
Yinghao Aaron LiCong HanNima MesgaraniPublished in: CoRR (2022)
Keyphrases
- generative model
- text to speech synthesis
- probabilistic model
- mixture model
- bayesian framework
- text to speech
- posterior probability
- discriminative learning
- em algorithm
- discriminative models
- prior knowledge
- latent dirichlet allocation
- expectation maximization
- markov chain monte carlo
- co occurrence
- semi supervised
- object categories
- feature representations
- scale space
- multiscale
- dirichlet process mixture models