Voice cloning is a speech synthesis technique that reproduces a specific person's timbre, intonation, and speaking style from recorded samples of their voice. Because the cloned voice can be used to read arbitrary text, it is considered a branch of text-to-speech (TTS). Early systems required hours of recordings and a separately trained model for each speaker, but with advances in deep learning-based speech synthesis, zero-shot approaches that clone a voice from only a few seconds to tens of seconds of audio have become common.
Voice cloning is used for video narration, dubbing, audiobooks, voice assistants, and assistive technology for people who have lost their voice to illness. Because cloning a voice without consent can be misused for fraud or disinformation, safeguards such as speaker consent verification and labeling of synthetic speech are discussed alongside the technology.