voice cloning
In plain wordsUsing AI to copy a specific person's voice from recordings, so a computer can say new things that sound like them — useful for dubbing and accessibility, and abused in scams.
The synthesis of new speech in the voice of a particular person from recordings of that person. Speaker-adaptive text-to-speechtext-to-speech, n.: Technology that reads written text aloud in a computer-generated voice, used in screen readers, voice assistants, audiobooks and navigation apps. systems either fine-tune on minutes to hours of the target's speech or, in zero-shot cloning, condition on a short reference clip through a speaker embedding or an audio prompt. Since 2023 leading systems treat speech as sequences of discrete tokens from a neural audio codecneural audio codec, n.: A neural network that squeezes sound into a very compact stream of numbers and rebuilds it, used both to send audio cheaply and to let AI models 'write' speech and music piece by piece. and continue a few seconds of reference audio with a language model, as in Microsoft's VALL-E, which could imitate a new speaker from a three-second recording, carrying over the speaker's emotional tone and the recording's acoustic conditions; diffusion and flow-matching decoders are also common. Voice conversion is the related task of transforming one person's speech so it sounds like another's. Legitimate uses include restoring voices lost to illness, dubbing and accessibility; misuse includes fraud (impersonated relatives or executives), political disinformation and non-consensual impersonation. Defenses include consent verification, watermarkingAI watermarking, n.: Hiding an invisible signal in AI-generated text, images, audio or video so that software can later tell the content was made by AI — like a banknote's watermark, but undetectable to the eye., detectiondeepfake detection, n.: Using software, often itself AI, to tell whether a video, photo or voice recording has been faked or generated by AI rather than captured from reality. and laws on digital replicasdigital replica, n.: A realistic computer-made copy of a real person's voice or appearance — such as an AI clone of a singer's voice or a digital stand-in for an actor — that can say or do things the person never did..
The bank now requires a callback because voice cloning makes a familiar voice proof of nothing.
Origin
Adapting speech synthesizers to new speakers predates deep learning. Neural few-shot cloning appeared around 2018 (Sercan Arık and colleagues at Baidu, 'Neural Voice Cloning with a Few Samples'); consumer services spread from 2023, the year Microsoft's VALL-E popularized codec-language-model cloning. A widely reported 2019 fraud used a cloned executive's voice to extract €220,000 from a UK energy firm.
Usage note
In February 2024 the US Federal Communications Commission ruled that AI-generated voices in robocalls fall under existing restrictions, after cloned-voice calls imitating President Joe Biden targeted New Hampshire primary voters. Consent rules for digital replicasdigital replica, n.: A realistic computer-made copy of a real person's voice or appearance — such as an AI clone of a singer's voice or a digital stand-in for an actor — that can say or do things the person never did. are spreading across jurisdictions.
Related
- text-to-speechtext-to-speech, n.: Technology that reads written text aloud in a computer-generated voice, used in screen readers, voice assistants, audiobooks and navigation apps.
- neural audio codecneural audio codec, n.: A neural network that squeezes sound into a very compact stream of numbers and rebuilds it, used both to send audio cheaply and to let AI models 'write' speech and music piece by piece.
- deepfakedeepfake, n.: A fake but realistic video, photo or voice recording made with AI that puts words in a real person's mouth or shows them in a scene that never happened.
- speaker verificationspeaker verification, n.: Checking whether a voice really belongs to the person it claims to be, as when a bank confirms your identity by the sound of your voice on the phone.
- digital replicadigital replica, n.: A realistic computer-made copy of a real person's voice or appearance — such as an AI clone of a singer's voice or a digital stand-in for an actor — that can say or do things the person never did.
- deepfake detectiondeepfake detection, n.: Using software, often itself AI, to tell whether a video, photo or voice recording has been faked or generated by AI rather than captured from reality.
- AI watermarkingAI watermarking, n.: Hiding an invisible signal in AI-generated text, images, audio or video so that software can later tell the content was made by AI — like a banknote's watermark, but undetectable to the eye.
- speaker verificationspeaker verification, n.: Checking whether a voice really belongs to the person it claims to be, as when a bank confirms your identity by the sound of your voice on the phone.
Sources
- Wang, C. et al. (2023). Neural Codec Language Models are Zero-Shot Text to Speech Synthesizers (VALL-E). arXiv:2301.02111. arxiv.org/abs/2301.02111
- Arık, S. Ö., Chen, J., Peng, K., Ping, W. & Zhou, Y. (2018). Neural Voice Cloning with a Few Samples. NeurIPS 2018.
Mentioned in
deepfakedeepfake, n.: A fake but realistic video, photo or voice recording made with AI that puts words in a real person's mouth or shows them in a scene that never happened., misusemisuse, n.: People deliberately using AI to cause harm — for scams, harassment, propaganda, hacking or weapons — as opposed to harm from AI making mistakes or pursuing goals of its own., text-to-speechtext-to-speech, n.: Technology that reads written text aloud in a computer-generated voice, used in screen readers, voice assistants, audiobooks and navigation apps., deepfake detectiondeepfake detection, n.: Using software, often itself AI, to tell whether a video, photo or voice recording has been faked or generated by AI rather than captured from reality., digital replicadigital replica, n.: A realistic computer-made copy of a real person's voice or appearance — such as an AI clone of a singer's voice or a digital stand-in for an actor — that can say or do things the person never did., music generationmusic generation, n.: Using AI to compose or produce music — from melodies written as notes to complete songs with vocals — often from a short text description like 'upbeat jazz with saxophone'., NO FAKES ActNO FAKES Act, prop. n.: A proposed US law that would let people control AI-made copies of their voice and face, and would require websites to take down unauthorized fakes once notified., right of publicityright of publicity, n.: A person's legal power to decide whether businesses may use their name, face, voice or other recognizable traits to sell things, and to be paid when they do. AI's skill at copying voices and faces makes it more important., speaker verificationspeaker verification, n.: Checking whether a voice really belongs to the person it claims to be, as when a bank confirms your identity by the sound of your voice on the phone., speech-to-speech modelspeech-to-speech model, n.: An AI model that listens to speech and replies or translates directly in speech, without first writing everything down as text, so conversations feel faster and keep the tone of voice., SunoSuno, prop. n.: A popular AI service that writes and performs complete songs — music, lyrics and singing — from a short text description., synthetic mediasynthetic media, n.: Images, video, audio or text created or substantially altered by AI rather than recorded or written directly by people, whether for art, entertainment, business or deception., voice agentvoice agent, n.: An AI system you talk to out loud, for example on a phone call, that listens, answers in a natural-sounding voice and can carry out tasks such as booking appointments or checking an order., Heart on My SleeveHeart on My Sleeve, prop. n.: A 2023 song that used AI to imitate the voices of Drake and The Weeknd without their involvement; it went viral before being pulled, becoming a landmark case for AI voice cloning in music., neural audio codecneural audio codec, n.: A neural network that squeezes sound into a very compact stream of numbers and rebuilds it, used both to send audio cheaply and to let AI models 'write' speech and music piece by piece.