Voice cloning consent statement example: Azure, Google, OpenAI

A consent statement for voice cloning is a short recording in which the speaker says that they own the voice and agree to a named party making a synthetic version of it. Azure and Google publish exact wording, OpenAI requires one of 16 fixed phrases, and all three check that the consent speaker is the person in the sample. Sume collects none of this, so a Sume clone needs a release you write and keep.

The three wordings

Azure's English (US) text, from Microsoft Learn, names the speaker and the company. Google's, from its Chirp 3 page, is a fixed sentence about owning the voice. OpenAI's guide requires one of its supported phrases, and any divergence fails.

What they have in common

Those four features are worth copying even when your tool does not enforce them.

• The speaker says it, in their own voice, in the language of the sample.
• The consent is recorded separately from the training sample.
• The vendor compares the consent voice with the sample voice.
• The text names who may use the voice: a company, Google, or the customer.

What a Sume voice needs from you

Sume's Voices library in the app has a create dialog that takes a name, a gender, a language and an uploaded or microphone-recorded clip. That dialog has no consent-recording step, and the public API has no route that creates a clone.

So write your own release. Name the speaker, the project, the length of the licence, where synthetic audio may appear, and how consent can be withdrawn. Record the speaker reading it, and file the audio with the voice id. For the law around AI voices, see the posts on the No Fakes Act and Tennessee's ELVIS Act.

Limits of this post

The wording above is each vendor's own and is for their tools only. It is not legal advice, and a vendor's statement does not satisfy another vendor's check. Re-read the pages before use, since they change. The Sume route is explained in Voice cloning API.