Google Inc.

Robert Andrew James Clark

Chun-an Chan

Ioannis Agiomyrgiannakis

Vincent Ping Leung Wan

Quoc V Le

Yonghui Wu

Ron J Weiss

Zhifeng Chen

Navdeep Jaitly

Ryan Michael Rifkin

Samuel Bengio

Russell J Ryan

Yuxuan Wang

Zongheng Yang

Ying Xiao

Yu Zhang

Byungha Chun

Vincent Wan

Ye Jia

Lev Finkelstein

Norman Casagrande

Byung Ha Chun

Tom Marius Kenter

Manish Kumar Sharma

Jakub Vit

Javier Gonzalvo

Aliaksei Severyn

Chun-An Chan

A method includes receiving an input text utterance to be synthesized into expressive speech having an intended prosody and a target voice and generating, using a first text-to-speech (TTS) model, an intermediate synthesized speech representation for the input text utterance. The intermediate synthesized speech representation possesses the intended prosody. The method also includes providing the intermediate synthesized speech representation to a second TTS model that includes an encoder portion and a decoder portion. The encoder portion is configured to encode the intermediate synthesized speech representation into an utterance embedding that specifies the intended prosody. The decoder portion is configured to process the input text utterance and the utterance embedding to generate an output audio signal of expressive speech that has the intended prosody specified by the utterance embedding and speaker characteristics of the target voice.

Two-level speech prosody transfer

A method includes obtaining training data including a plurality of training audio signals and corresponding transcripts. Each training audio signal is spoken by a target speaker in a first accent/dialect. For each training audio signal of the training data, the method includes generating a training synthesized speech representation spoken by the target speaker in a second accent/dialect different than the first accent/dialect and training a text-to-speech (TTS) system based on the corresponding transcript and the training synthesized speech representation. The method also includes receiving an input text utterance to be synthesized into speech in the second accent/dialect. The method also includes obtaining conditioning inputs that include a speaker embedding and an accent/dialect identifier that identifies the second accent/dialect. The method also includes generating an output audio waveform corresponding to a synthesized speech representation of the input text sequence that clones the voice of the target speaker in the second accent/dialect.

Two-level text-to-speech systems using synthetic training data

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating speech from text. One of the systems includes one or more computers and one or more storage devices storing instructions that when executed by one or more computers cause the one or more computers to implement: a sequence-to-sequence recurrent neural network configured to: receive a sequence of characters in a particular natural language, and process the sequence of characters to generate a spectrogram of a verbal utterance of the sequence of characters in the particular natural language; and a subsystem configured to: receive the sequence of characters in the particular natural language, and provide the sequence of characters as input to the sequence-to-sequence recurrent neural network to obtain as output the spectrogram of the verbal utterance of the sequence of characters in the particular natural language.

End-to-end text-to-speech conversion

A method for generating a prosodic representation includes receiving a text utterance having one or more words. Each word has at least one syllable having at least one phoneme. The method also includes generating, using a Bidirectional Encoder Representations from Transformers (BERT) model, a sequence of wordpiece embeddings and selecting an utterance embedding for the text utterance, the utterance embedding representing an intended prosody. Each wordpiece embedding is associated with one of the one or more words of the text utterance. For each syllable, using the selected utterance embedding and a prosody model that incorporates the BERT model, the method also includes generating a corresponding prosodic syllable embedding for the syllable based on the wordpiece embedding associated with the word that includes the syllable and predicting a duration of the syllable by encoding linguistic features of each phoneme of the syllable with the corresponding prosodic syllable embedding for the syllable.

Speech synthesis prosody using a BERT model

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training a machine learning model to generate embeddings of inputs to the machine learning model, the machine learning model having an encoder that generates the embeddings from the inputs and a decoder that generates outputs from the generated embeddings, wherein the embedding is partitioned into a sequence of embedding partitions that each includes one or more dimensions of the embedding, the operations comprising: for a first embedding partition in the sequence of embedding partitions: performing initial training to train the encoder and a decoder replica corresponding to the first embedding partition; for each particular embedding partition that is after the first embedding partition in the sequence of embedding partitions: performing incremental training to train the encoder and a decoder replica corresponding to the particular partition.

Training neural networks to generate structured embeddings

A method of providing a frame-based mel spectral representation of speech includes receiving a text utterance having at least one word and selecting a mel spectral embedding for the text utterance. Each word has at least one syllable and each syllable has at least one phoneme. For each phoneme, the method further includes using the selected mel spectral embedding to: (i) predict a duration of the corresponding phoneme based on corresponding linguistic features associated with the word that includes the corresponding phoneme and corresponding linguistic features associated with the syllable that includes the corresponding phoneme; and (ii) generate a plurality of fixed-length predicted mel-frequency spectrogram frames based on the predicted duration for the corresponding phoneme. Each fixed-length predicted mel-frequency spectrogram frame represents mel-spectral information of the corresponding phoneme.

Clockwork hierarchal variational encoder

A method includes receiving an input text utterance to be synthesized into expressive speech having an intended prosody and a target voice and generating, using a first text-to-speech (TTS) model, an intermediate synthesized speech representation tor the input text utterance. The intermediate synthesized speech representation possesses the intended prosody. The method also includes providing the intermediate synthesized speech representation to a second TTS model that includes an encoder portion and a decoder portion. The encoder portion is configured to encode the intermediate synthesized speech representation into an utterance embedding that specifies the intended prosody. The decoder portion is configured to process the input text utterance and the utterance embedding to generate an output audio signal of expressive speech that has the intended prosody specified by the utterance embedding and speaker characteristics of the target voice.

A method for providing a frame-based mel spectral representation of speech includes receiving a text utterance having at least one word, and selecting a mel spectral embedding for the text utterance. Each word in the text utterance has at least one syllable and each syllable has at least one phoneme. For each phoneme, using the selected mel spectral embedding, the method also includes: predicting a duration of the corresponding phoneme by encoding linguistic features of the corresponding phoneme with a corresponding syllable embedding for the syllable that includes the corresponding phoneme; and generating a plurality of fixed-length predicted mel-frequency spectrogram frames based on the predicted duration for the corresponding phoneme. Each fixed-length predicted mel-frequency spectrogram frame representing mel-spectral information of the corresponding phoneme.

Clockwork hierarchical variational encoder

Methods, systems, and computer-readable media for text-to-speech synthesis using an autoencoder. In some implementations, data indicating a text for text-to-speech synthesis is obtained. Data indicating a linguistic unit of the text is provided as input to an encoder. The encoder is configured to output speech unit representations indicative of acoustic characteristics based on linguistic information. A speech unit representation that the encoder outputs is received. A speech unit is selected to represent the linguistic unit, the speech unit being selected from among a collection of speech units based on the speech unit representation output by the encoder. Audio data for a synthesized utterance of the text that includes the selected speech unit is provided.

Text-to-speech synthesis using an autoencoder

Growing community of inventors

Hertfordshire, United Kingdom

Robert Andrew James Clark