Descript’s AI Speakers let you create audio with ultra-realistic voices—either using a custom clone of your own voice or one of Descript’s stock voices. You can use AI Speakers to write narration, fix mistakes, correct tone, and fine-tune your audio without re-recording.

With AI Speakers, you can

• Create a custom AI Speaker
• Generate text-to-speech
• Fix or rewrite audio with Regenerate
• Convert AI clips to editable audio

Create a custom AI Speaker

Create an AI Speaker in your own voice using as little as 30 seconds of audio. Once trained, your voice can be used to generate new content or fix mistakes without re-recording. Learn how to create a custom AI Speaker.

Generate text-to-speech

Type directly into the script and select an AI Speaker to narrate your words. You can use your own cloned voice or one of Descript’s ultra-realistic stock voices. Learn how to generate text-to-speech.

Fix or rewrite audio with Regenerate

The Regenerate feature lets you rewrite individual words or phrases using your AI voice—perfect for fixing errors or updating phrasing without re-recording. Learn more about Descript's Regenerate features.

Convert AI clips to editable audio

AI-generated clips behave differently from organic recorded audio. You’ll need to convert them before you can apply fades, speed changes, or timeline edits.

To convert a clip:

1. Hover over the AI voice clip in the timeline.
2. Click the Convert to audio button that appears.

Once converted, the clip behaves like any standard audio file in Descript.

Adjust tone and style of your AI Speaker

Already created a custom AI Speaker but want a different delivery? You can now guide tone and style in two ways:

• Create a new AI Speaker by recording a sample in the tone you want. The voice sample you provide determines how the model sounds.
• Or, for non-legacy subscribers, use tags and inline prompts with the ElevenLabs v3 model to adjust delivery directly. Learn more about the v3 model