Who can create an Overdub Voice / What is a Consenting Speaker?

Who can Create a custom Overdub Voice?

From section 9.3 of the Descript Terms of Service:

Overdub is a feature of Descript that allows you to use Overdub Voices to synthesize voice recordings that sound like a Consenting Speaker merely by typing the new words... A “Consenting Speaker” is a person (including you) that has given explicit consent to use recordings of their voice to train an Overdub Voice and to synthesize and otherwise use their voice.


How must a Consenting Speaker grant authorization to create their voice?

From the "Overdub Voices" section of the Descript Security and Confidentiality page:

You may only use recordings of 'Consenting Speakers' to create 'Overdub Voices.' As part of the Overdub Voice creation process, Speakers must read a script, designated by Descript, which states the Speaker’s identity and affirms their consent to Overdub synthesizing their voice. No other voice recording can be used for creation of an Overdub Voice.


Can someone else use Overdub to clone my voice without my permission?

When recording Overdub Voice training data, Descript users must positively affirm their identity by recording a Voice ID statement and express consent to generating their Overdub synthesized voice.

We verify this consent by authenticating the audio uploaded against our training script to ensure that the voice recorded is consenting to voice synthesis.

Training data which does not include this Voice ID cannot be used to create an Overdub Voice.


Can I create an Overdub Voice for a deceased individual?

For security reasons, we cannot create an Overdub Voice without an explicit Voice ID statement recording from a Consenting Speaker. As such, we cannot create an Overdub Voice for a deceased person.


I am currently unable to speak, or am acting on behalf of an individual who is unable to speak. Can I use Overdub to create a digital copy of a voice using existing recordings?

No. For security reasons, we cannot create an Overdub Voice without an explicit verbal consent recording from the Consenting Speaker.


Due to circumstances beyond my control, I am unable to provide Voice ID audio that sufficiently matches my training data. Can I provide other forms of Speaker Consent verification in order to receive an exception to this process?

Your Voice ID consent statement will be analyzed and compared against the provided training audio to assure that they match. If there is insufficient matched audio, your voice training will fail and you will receive an email notification from Descript that the training has failed due to insufficient matched audio. As such, Descript cannot accept other forms of voice verification or Speaker Consent beyond the current matching technology.