Google's Translatotron is an end-to-end model that mimics human voices

Google AI today introduced Translatotron, an end-to-end speech-to-speech model capable of retaining the sound of the voice that's being translated.