Voice Imitation Technology allows anyone to customize any Text To Speech (TTS) voice with one, or several, Target Voices. This process requires the recording, in a studio, of a corpus of about a hundred sentences made by the Target Voice interpreter, or the choice of a Target Voice from archive audio files.
CandyVoice creates the supervised voice model of the Target Voice from the studio recordings, or from the archive audio files. The quality of the voice model depends of the recording quality of the Target Voice (sentences pronunciation and acoustic environment), or, if applicable, of the archive audio files’ quality.
To customize the TTS, the user sends a text file to be synthesized with the Target Voice via CandyVoice’s API. Then, almost instantly, he receives an audio file where the text is read with the Target Voice. The quality of the final result depends not only of the quality of the voice model of the Target Voice, but also of the quality of the TTS’s voice.
The voice imitation technology which allows anyone to imitate multiple Target Voices by multiple Sources Voices works also in real time. This process requires the recording in the studio of a corpus of about hundred sentences by the interpreters of the Source Voice and the Target Voice. Target Voice can also come from an archive audio file.
CandyVoice creates supervised voice models of Source and Target Voices from studio recordings, or from an archive audio file. The quality of voice models depends of the recording quality (sentences pronunciation and acoustic environment), or, if applicable, of the archive audio file’s quality.
CandyVoice’s technology allows to imitate in real-time multiple Target Voices (including the voices of personalities) by multiple Source Voices (and vice versa!). This technology find its use, for example, in entertainment industry and video games, where the player can animate the game’s character with his own voice in real time.
