Voice Cloning Consent Policy

Velluto Last updated: 4 июня 2026 г.

This Voice Cloning Consent Policy explains the rules and procedures that apply when you create a synthetic voice profile ("Voice Model") through the Services. It is part of the Terms of Service and is to be read together with the Acceptable Use Policy and the Privacy Policy.

Creating a Voice Model is one of the most powerful — and most sensitive — features of the Services. Misuse of voice cloning causes real harm to real people. This Policy is designed to make sure cloning is done only with informed, specific, and revocable consent of the voice owner.

1. Two types of voices

1.1. Curated Voices

A "Curated Voice" is a synthetic voice published in our voice catalogue and made available to all eligible users. We have obtained the necessary rights or licences to publish each Curated Voice. You may use Curated Voices according to the licence terms set out for each voice in the catalogue.

1.2. Personal Voice Models

A "Personal Voice Model" is a synthetic voice you create by uploading audio. Personal Voice Models are private to your Account unless you explicitly share them.

2. Eligibility: whose voice you may clone

You may upload audio to create a Personal Voice Model only if one of the following is true:

(a) Self. The audio is a recording of your own voice; or

(b) Verified third party. The audio is a recording of another individual ("Voice Owner") from whom you have obtained, before upload:

  • explicit — given in words, not implied;
  • specific — referring to the creation of a synthetic voice through Velluto;
  • informed — provided after being told what a Voice Model is, what it can do, and that it will be used commercially (where applicable);
  • revocable — the Voice Owner can withdraw at any time; and
  • written — captured in a document signed (electronically or otherwise) by the Voice Owner that you can produce on request.

For our professional tier ("Premium Voice Cloning") we accept own voice only. You may not clone any third-party voice through that tier, even with consent.

3. Voices you may never clone

You may not create or attempt to create a Voice Model of:

(a) any individual without satisfying Section 2;

(b) any minor (any person under 18 years of age) — including with parental consent;

(c) any deceased person without verified authorisation from the rights holder of their estate (provide a copy of the authorisation on request);

(d) any current or recent political candidate, elected official, judicial officer, or head of state — regardless of whether you claim to have obtained that person's authorisation. "Recent" means within five (5) years of the person leaving office or losing election;

(e) any celebrity, artist, musician, athlete, broadcaster, voice actor, or public figure whose voice is generally recognisable to the public, without verified written authorisation;

(f) any voice whose owner appears on our published No-Go Voices list (a list of persons who have notified us that they do not consent to being cloned);

(g) any voice for which you intend to produce content prohibited under the AUP, including non-consensual intimate imagery, fraud, defamation, impersonation, election manipulation, or harassment.

4. Verification at upload

When you create a Personal Voice Model you must complete the in-product Consent Gate, which requires you to:

4.1. Confirm by checkbox the following representations and warranties (this text is captured verbatim at upload time and stored against your Account):

I represent and warrant that:

(a) the voice in the audio I am uploading is either my own voice, or the voice of a Voice Owner from whom I have obtained explicit, specific, informed, and revocable written consent to create and commercially use a synthetic voice model through Velluto;

(b) I have the legal authority to grant Velluto the licences described in the Terms with respect to this audio;

(c) the voice does not belong to a minor, a deceased person without estate authorisation, a current or recent political candidate or elected official, or any person whose name appears on Velluto's No-Go Voices list;

(d) my use of any Voice Model generated will comply with the Acceptable Use Policy and all applicable laws, including the Tennessee ELVIS Act, the EU AI Act, California Civil Code §3344, and any other right-of-publicity or biometric-privacy law in the jurisdictions where the Output will be distributed;

(e) I will retain, for at least three (3) years, written evidence of consent in a form that I can produce on request within 14 days;

(f) I will indemnify Velluto per the Terms for any breach of these representations.

4.2. Voice captcha (for Premium Voice Cloning): we generate a unique phrase that must be read aloud within ten (10) seconds. We compare the resulting audio against your uploaded sample to confirm it is the same person.

4.3. Read-aloud consent statement (where applicable): we provide a specific phrase such as "I, [name], on [date], consent to the creation of a synthetic voice based on my recordings, to be used by Velluto." You must include a recording of the Voice Owner reading this phrase, time-stamped, before activation.

We may add or change verification steps to combat new abuse patterns.

5. Ownership, licences, and rights

5.1. Audio Output. Subject to your continued compliance with the Terms, you own the audio Output you generate using a Personal Voice Model.

5.2. Voice Model. The Voice Model itself is hosted by us on your behalf. You grant us a non-exclusive, worldwide, royalty-free licence to host, store, process, transmit, and use the audio and resulting model solely as necessary to provide the Services to you, to ensure security, and to comply with law.

5.3. Voice Owner — independent rights. A Voice Owner (the natural person whose voice the model is based on) retains, at any time and regardless of who uploaded the audio, the right to:

(a) request deletion of the Voice Model;

(b) request that we cease generating new Output from it; and

(c) request a copy of the audio originally uploaded in respect of their voice.

We honour verified Voice Owner requests within thirty (30) days of receipt, regardless of any contract between you and the Voice Owner. This right is independent of the right of erasure under data-protection law and operates as a standalone contractual right granted by us to every Voice Owner.

5.4. Effect of revocation. Where consent is revoked, the Voice Model is deleted. Output previously generated and already in your possession may continue to be used under the Terms and any applicable third-party licence, unless continued use would itself be unlawful (for example, if you have lost the licence to use the voice for the specific purpose in question).

5.5. No training without consent. We do not use your audio uploads or Voice Models to train general-purpose AI models. Aggregated, de-identified telemetry may be used to improve system performance.

6. Watermarking and detection

To support trust and to align with the EU AI Act (Article 50, effective 2 August 2026) and other emerging transparency rules, audio Output generated through the Services may carry machine-readable markers indicating that it is AI-generated. These markers are designed not to affect perceptual quality. Removing or attempting to remove them is a violation of the AUP.

7. Takedown procedure (for Voice Owners and third parties)

If you are the Voice Owner of a voice that has been cloned through the Services, or if you believe a Voice Model violates this Policy, contact:

[email protected]

Subject line: "Voice Cloning Takedown — [your name]"

Please include:

  1. Your full name and contact details.
  2. Confirmation that you are the Voice Owner (or authorised representative — please attach proof of authority).
  3. Description of the issue (unauthorised clone / abuse of consent / use prohibited under AUP).
  4. Any identifying information (Account email if known, voice name in the Services if known, audio sample of your real voice for matching).
  5. A statement, signed under penalty of perjury (or its equivalent in your jurisdiction): "I have a good-faith belief that I am the Voice Owner of the voice identified, and the use of my voice through Velluto has not been authorised by me."

We acknowledge takedowns within 72 hours and resolve verified cases within 30 days. We may, at our discretion, suspend the Voice Model immediately on receipt of a credible report and pending verification.

We cooperate fully with law enforcement in cases of voice-based fraud, sextortion, NCII, or election manipulation.

8. No-Go Voices

We maintain a list of voices that may not be cloned through the Services ("No-Go Voices"). The list is non-public to avoid providing a target catalogue, but voices added include:

  • voices of persons who have notified us of their non-consent;
  • voices of current and recent political candidates and elected officials;
  • voices of well-known public figures whose unauthorised cloning would cause demonstrable harm;
  • voices that have been the subject of repeated abuse reports.

If you believe you should be added to the No-Go Voices list, contact [email protected].

9. Sanctions for violation

Violation of this Policy is a material breach of the Terms. Consequences include, in our sole discretion:

  • immediate deletion of the Voice Model and any associated Output stored with us;
  • immediate suspension or termination of your Account;
  • forfeiture of prepaid fees;
  • reporting to law-enforcement authorities;
  • cooperation with civil claims by Voice Owners, including production of all records relating to your use of the Services in response to valid legal process;
  • public disclosure where required by law or warranted by the severity of the abuse.

You agree to indemnify us for any third-party claim arising from your violation of this Policy, per the Terms.

10. Contact

Questions about this document? [email protected].