VoiceGen

Low latency, on-prem / on-cloud solutions for highly natural streaming text to speech synthesis.

VoiceGen synthesizes speech from text. The API has a single synthesis call, StreamingSynthesize: you send the text and a SynthesisConfig naming the model, the speaker and the audio format you want, and the server streams audio back in chunks as it is produced — so playback can start before the whole utterance is synthesized.

ListModels reports which voices a server has loaded, including the speakers available in each model and the phone set it accepts, which is what you need if you plan to send pronunciations rather than plain text.

Like the other engines, VoiceGen runs on your own hardware, on prem, in your private cloud, or embedded.