What is voice cloning?

Voice cloning is AI technology that learns a specific person's voice from a short audio sample and can then generate new speech in that voice. With as little as 30 seconds to a few minutes of source audio, a modern voice-cloning system can produce new sentences that sound convincingly like the original speaker.

Voice cloning is a subset of generative voice AI — specifically, generative text-to-speech that has been conditioned on a particular voice rather than trained to produce a generic voice.

How voice cloning works

• Enrollment — the system is given a short sample of the target speaker's voice, typically 30 seconds to 10 minutes of clean audio

• Model conditioning — a generative voice model is conditioned on acoustic features extracted from the sample (pitch, timbre, cadence, vowel shape)

• Synthesis — the conditioned model generates new audio from any text input, preserving the original speaker's characteristics

Some cloning systems fine-tune a dedicated model per voice (higher quality, higher cost). Others use a single large model with "voice embeddings" that can switch between any enrolled voice on demand (faster, more flexible).

What voice cloning is used for

• Branded AI voice agents — a business hires a voice actor, records a sample, and deploys an AI receptionist that speaks in that specific voice as a brand asset

• Audiobook and podcast production — letting an author's voice narrate additional content without re-recording

• Assistive technology — preserving the voice of someone facing a medical condition that will affect their speech, or giving a personalized voice to someone who has lost theirs

• Localization — generating voice performances in additional languages while preserving the original speaker's identity

• Gaming and entertainment — generating character dialogue at scale in a consistent voice

The ethics and consent side

Voice cloning is powerful enough that it raises serious consent and misuse concerns. Responsible voice-cloning platforms require:

• Explicit consent from the person whose voice is being cloned

• Verification that the enrolling party has the right to use the voice (for commercial cloning)

• Watermarking or provenance signals so generated audio can be identified as synthetic

• Prohibitions on cloning public figures, politicians, or specific protected individuals without verified consent

• Terms of service limiting use to the enrolling party's own business

Before cloning a voice for commercial use, work with a platform that enforces these standards and ensures the voice you're cloning has authorized its use.

Related concepts

• Generative voice AI — the broader category

• Text-to-speech (TTS) — what voice cloning extends

• Voice AI — the overall space

See it in action

The Receptionist Agent at 365agents supports voice cloning for customers who want to brand their AI with a custom voice. We require written consent from the voice subject and verify ownership before enabling a cloned voice in production. Contact us for the voice-cloning intake process.