A computer system is provided that includes a processor configured to store a set of audio training data that includes a plurality of audio segments and metadata indicating a word or phrase associated with each audio segment. For a target training statement of a set of structured text data, the processor is configured to generate a concatenated audio signal that matches a word content of a target training statement by comparing the words or phrases of a plurality of text segments of the target training statement to respective words or phrases of audio segments of the stored set of audio training data, selecting a plurality of audio segments from the set of audio training data based on a match in the words or phrases between the plurality of text segments of the target training statement and the selected plurality of audio segments, and concatenating the selected plurality of audio segments.

A method, computer program, and computer system is provided for converting a singing first singing voice associated with a first speaker to a second singing voice associated with a second speaker. A context associated with one or more phonemes corresponding to the first singing voice is encoded, and the one or more phonemes are aligned to one or more target acoustic frames based on the encoded context. One or more mel-spectrogram features are recursively generated from the aligned phonemes and target acoustic frames, and a sample corresponding to the first singing voice is converted to a sample corresponding to the second singing voice using the generated mel-spectrogram features.

A method, computer program, and computer system is provided for converting a singing first singing voice associated with a first speaker to a second singing voice associated with a second speaker. A context associated with one or more phonemes corresponding to the first singing voice is encoded, and the one or more phonemes are aligned to one or more target acoustic frames based on the encoded context. One or more mel-spectrogram features are recursively generated from the aligned phonemes and target acoustic frames, and a sample corresponding to the first singing voice is converted to a sample corresponding to the second singing voice using the generated mel-spectrogram features.

Generation of synthetic speech from an input text sequence may be difficult when durations of individual phonemes forming the input text sequence are unknown. A predominantly parallel process may model speech rhythm as a separate generative distribution such that phoneme duration may be sampled at inference. Additional information such as pitch or energy may also be sampled to provide improved diversity for synthetic speech generation.

A method, computer program, and computer system is provided for converting a singing first singing voice associated with a first speaker to a second singing voice associated with a second speaker. A context associated with one or more phonemes corresponding to the first singing voice is encoded, and the one or more phonemes are aligned to one or more target acoustic frames based on the encoded context. One or more mel-spectrogram features are recursively generated from the aligned phonemes and target acoustic frames, and a sample corresponding to the first singing voice is converted to a sample corresponding to the second singing voice using the generated mel-spectrogram features.

A method, computer program, and computer system is provided for converting a singing first singing voice associated with a first speaker to a second singing voice associated with a second speaker. A context associated with one or more phonemes corresponding to the first singing voice is encoded, and the one or more phonemes are aligned to one or more target acoustic frames based on the encoded context. One or more mel-spectrogram features are recursively generated from the aligned phonemes and target acoustic frames, and a sample corresponding to the first singing voice is converted to a sample corresponding to the second singing voice using the generated mel-spectrogram features.

A method for disentangled variational speech representation learning for voice conversion, performed by at least one processor, is provided. The method includes receiving input speech segments, encoding the input speech segments via a shared encoder to generate a speaker embedding and a content embedding, encoding the posterior distributions of the speaker embedding via a speaker encoder and encoding the posterior distributions of the content embedding via a content encoder to obtain encoded results, and decoding the encoded results by concatenating the encoded results to obtain a reconstructed speech output.

A method for disentangled variational speech representation learning for voice conversion, performed by at least one processor, is provided. The method includes receiving input speech segments, encoding the input speech segments via a shared encoder to generate a speaker embedding and a content embedding, encoding the posterior distributions of the speaker embedding via a speaker encoder and encoding the posterior distributions of the content embedding via a content encoder to obtain encoded results, and decoding the encoded results by concatenating the encoded results to obtain a reconstructed speech output.

Devices and techniques are generally described for learned condition text-to-speech synthesis. In some examples, first data representing a selection of a type of prosodic expressivity may be received. In some further examples, a selection of content comprising text data may be received. First audio data may be determined that includes an audio representation of the text data. The first audio data may be generated based at least in part on sampling from a first latent distribution generated using a conditional primary variational autoencoder (VAE). The sampling from the first latent distribution may be conditioned on a first learned distribution associated with the type of prosodic expressivity. In various examples, the first audio data may be sent to a first computing device.

Devices and techniques are generally described for learned condition text-to-speech synthesis. In some examples, first data representing a selection of a type of prosodic expressivity may be received. In some further examples, a selection of content comprising text data may be received. First audio data may be determined that includes an audio representation of the text data. The first audio data may be generated based at least in part on sampling from a first latent distribution generated using a conditional primary variational autoencoder (VAE). The sampling from the first latent distribution may be conditioned on a first learned distribution associated with the type of prosodic expressivity. In various examples, the first audio data may be sent to a first computing device.