Emotion-based text to speech

A system comprising:

• at least one processor configured to perform operations comprising:
• storing a first embedding for a first speaker of a plurality of speakers and a second embedding for a second speaker of the plurality of speakers;
• selecting, by one or more processors, the first speaker for virtual speech of one or more words of a text string;
• generating, based on the first embedding of the first speaker and the second embedding of the second speaker, a third embedding for the first speaker;
• and
• applying the third embedding of the first speaker and the text string to a vocoder to generate an audio stream comprising the one or more words being virtually spoken by the first speaker.

A non-transitory machine-readable storage medium that includes instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:

• storing a first embedding for a first speaker of a plurality of speakers and a second embedding for a second speaker of the plurality of speakers;
• selecting, by one or more processors, the first speaker for virtual speech of one or more words of a text string;
• generating, based on the first embedding of the first speaker and the second embedding of the second speaker, a third embedding for the first speaker;
• and
• applying the third embedding of the first speaker and the text string to a vocoder to generate an audio stream comprising the one or more words being virtually spoken by the first speaker.