AI voice cloning from 10 seconds of audio

AI voice cloning creates a synthetic voice from recordings of a speaker. You can then generate new speech from text using that voice. The result depends on the source recording and the cloning method.

Start with 10 seconds for an Instant Voice Clone. Sonic 3.6 and newer can use up to 60 seconds to better retain the speaker's accent; older models only learn from the first 10 seconds. Check the upload form for its current file-size limit. A Pro Voice Clone requires at least 30 minutes of clean audio from one speaker. See the current Instant Voice Clone and Pro Voice Clone recording guides.

Instant Voice Cloning creates a voice in seconds from a short clip. Pro Voice Cloning trains on a larger dataset and is worth trying when an instant clone does not preserve a rare accent or distinctive voice qualities closely enough. Pro Voice Clone training takes up to 3 hours according to the current recording guide. Its speed and volume are learned from the dataset and cannot be changed with request-time controls.

Record one speaker in a quiet room without music, echo, or background noise. Speak naturally in the language and style you want the clone to use. For an instant clone, the mood of your clip carries into the generated voice. For a pro clone, approve a test read before recording the full dataset and include the types of phrases your agent will actually say.

Yes. Open Instant Voice Clone in the Cartesia Playground, record or upload your voice, and choose the language of the recording. You can then enter a script and listen to your custom voice read it. You do not need to build an API integration to get started.

You can try cloning free with the demo on this page. To keep a clone in your account, Instant Voice Cloning starts on Pro and Pro Voice Cloning on Startup. Compare the current plans.

Yes. A voice clone gives you a reusable voice for generating speech from new text. Use it to narrate a tutorial, revise a video script, or read a presentation in your own voice. Listen to the generated audio before adding it to your project, especially when the script contains names or specialist terms.

Usually. The clone takes its accent and delivery from your recording. Sonic 3.6 and newer can learn from up to 60 seconds to hold an accent more closely. For a rare accent, a Pro Voice Clone trains on 30 minutes or more.

Create the voice through the Playground or the cloning API, then use its voice ID in your text-to-speech requests. Your application supplies the text to speak and receives generated audio. Start with the Clone Voice API reference for the recording upload and voice-creation steps.

Yes, you can use a custom voice in a Cartesia voice agent. Voice cloning supplies the voice; the agent still needs instructions, business-system tools, and a phone connection. Our AI receptionist guide explains scheduling, human handoffs, and tests to run before taking customer calls.

Only clone a voice you own or have explicit permission to clone and use for the intended purpose. A subscription does not grant rights to another person's voice. Check your plan's access and commercial-use terms on the pricing page and review the terms of service before deployment.