Overview
A custom voice is a consent backed synthetic copy of your own voice, built from a short recording you provide. Once enrolled, it can narrate Studio videos in your voice instead of a catalogue voice.
Enrollment goes through Google Cloud or OpenAI's voice cloning services. Raw recordings are processed in memory and discarded after enrollment; only an encrypted provider credential remains.
Before You Enroll
You must own the voice you are enrolling and have the legal right to use it for AI synthesis. Enrollment requires two short recordings:
- Reference recording: 8 to 10 seconds of natural speech in a quiet room.
- Consent recording: a short spoken statement confirming you own the voice and consent to its use, worded by the provider you choose.
One Voice Captures One Energy
The reference recording sets the energy your voice defaults to. Read it with the delivery you want this voice to use day to day: calm, upbeat, formal, whatever fits most of your narration.
A single enrolled voice cannot be told to sound excited on one line and calm on the next. There is no setting that changes a cloned voice's emotion or volume after enrollment. If you want a noticeably different delivery, such as an energetic version alongside a calm one, enroll a second voice profile with its own reference recording read in that register, then give each profile its own name so you can tell them apart.
Enrolling a Voice
- Open workspace settings and find Your synthetic voice.
- Name the profile and choose a provider.
- Record or upload the reference recording, reading it with the delivery you want this profile to default to.
- Record or upload the consent recording, reading the exact statement shown.
- Confirm you own the voice, then create and activate it.
A profile can be previewed, deactivated, or deleted at any time from the same screen.
Using a Custom Voice
Once active, a custom voice becomes available as a narration voice in Studio. For a script with more than one speaker, an authored video prompt can name each enrolled profile under its own speaker and assign it to specific scenes, so different registers can appear in the same video.