Abstract: A speech model includes a sub-model corresponding to a vocal attribute. The speech model generates an output waveform using a sample model, which receives text data, and a conditioning model, which receives text metadata and produces a prosody output for use by the sample model. If, during training or runtime, a different vocal attribute is desired or needed, the sub-model is re-trained or switched to a different sub-model corresponding to the different vocal attribute.

Abstract: A speech model is trained using multi-task learning. A first task may correspond to how well predicted audio matches training audio; a second task may correspond to a metric of perceived audio quality. The speech model may include, during training, layers related to the second task that are discarded at runtime.

Abstract: Wideband speech is synthesized from a bandlimited speech signal, for example from speech which has been transmitted via the public switched telephone network. Due to the nature of the vocal tract, there is a correlation between a bandlimited signal and those parts of an original wideband speech signal which are missing from that signal. Narrowband speech is characterized in terms of estimated formant frequencies provided by a peak picker. The frequency of formants in speech give a good indication, for voiced sounds, as to the shape of the vocal tract. The set of frequencies provided by the peak picker is used to access a codebook which provides synthesis parameters for use by a synthesizer.

Abstract: This invention relates to the generation of synthetic speech and specifically to the production of a digital waveform from a text in phonemes. The invention uses a linked database which comprises an extended text in phonemes and its equivalent in the form of a digital waveform. The two portions of the database are linked by a parameter which establishes equivalent points in both the phoneme text and the digital waveform. The input text (in phonemes) is analyzed to locate matching portion in the phoneme portion of the database. This matching utilises exact equivalence of phonemes where this is possible; otherwise relation between phonemes is utilised. The selection process identifies input phonemes in context whereby improved conversions are obtained. Having analyzed the input text into matching strings in the input form of the database beginning and ending parameters for the sections are established.

Abstract: Synthetic speech is generated by production of a digital waveform from a text in phonemes. A linked database is used which comprises an extended text in phonemes and its equivalent in the form of a digital waveform. The two portions of the database are linked by a parameter which establishes equivalent points in both the phoneme text and the digital waveform. The input text (in phonemes) is analyzed to locate a matching portion in the phoneme portion of the database. This matching utilizes exact equivalence of phonemes where this is possible; otherwise relation between phonemes is utilized. The selection process identifies input phonemes in context whereby improved conversions are obtained. Having analyzed the input exit into matching strings in the input form of the database beginning and ending parameters for the sections are established. The output text is produced by abutting sections of the digital waveform and defined by the beginning and ending parameters.

Abstract: Portions of recorded speech waveform (e.g., corresponding to phonemes) are combined to synthesize words. In order to provide a smoother delivery, each voiced portion of a waveform portion has its amplitude adjusted to a predetermined reference level. The scaling factor used is varied gradually over a transition region between such portions and between voiced and unvoiced portions.

Abstract: Synthetic speech is generated by production of a digital waveform from a text in phonemes. A linked database is used which comprises an extended text in phonemes and its equivalent in the form of a digital waveform. The two portions of the database are linked by a parameter which establishes equivalent points in both the phoneme text and the digital waveform. The input text (in phonemes) is analyzed to locate a matching portion in the phoneme portion of the database. This matching utilizes exact equivalence of phonemes where this is possible; otherwise relation between phonemes is utilized. The selection process identifies input phonemes in context whereby improved conversions are obtained. Having analyzed the input exit into matching strings in the input form of the database beginning and ending parameters for the sections are established. The output text is produced by abutting sections of the digital waveform and defined by the beginning and ending parameters.