Using previous acoustic context to improve Text-to-Speech synthesis

7 Dec 2020 • Pilar Oplustil-Gallegos • Simon King

Many speech synthesis datasets, especially those derived from audiobooks, naturally comprise sequences of utterances. Nevertheless, such data are commonly treated as individual, unordered utterances both when training a model and at inference time... (read more)

PDF Abstract

Results from the Paper

Add Remove

Submit results from this paper to get state-of-the-art GitHub badges and help the community compare results to other papers.

METHOD	TYPE
Sigmoid Activation	Activation Functions
LSTM	Recurrent Neural Networks
BiLSTM	Bidirectional Recurrent Neural Networks
Dilated Causal Convolution	Temporal Convolutions
Zoneout	Regularization
WaveNet	Generative Audio Models
Linear Layer	Feedforward Networks
Residual GRU	Recurrent Neural Networks
Highway Layer	Miscellaneous Components
Residual Connection	Skip Connections
Batch Normalization	Normalization
Mixture of Logistic Distributions	Output Functions
Convolution	Convolutions
Dense Connections	Feedforward Networks
Location Sensitive Attention	Attention Mechanisms
Max Pooling	Pooling Operations
GRU	Recurrent Neural Networks
BiGRU	Bidirectional Recurrent Neural Networks
Tanh Activation	Activation Functions
Highway Network	Feedforward Networks
Dropout	Regularization
Tacotron 2	Text-to-Speech Models
ReLU	Activation Functions
CBHG	Speech Synthesis Blocks
Griffin-Lim Algorithm	Phase Reconstruction
Additive Attention	Attention Mechanisms
Tacotron	Text-to-Speech Models

METHOD

TYPE

Sigmoid Activation