Voice cloning consent form: what to include
Training a voice from someone’s recording is only fair, and in many places only legal, when that person clearly agreed. A short written consent protects the speaker, protects you and makes later questions easy to answer: a client asks where the voice came from, a platform asks for proof, the speaker changes their mind.
This guide lists what a voice consent form should cover, in plain words. It is a practical checklist, not legal advice; for contracts that matter, ask a lawyer in your country.
Why written consent matters
Never clone a voice from recordings found online, from a call or from a video without the person’s agreement.
- The speaker knows exactly what will be made from their voice.
- You can prove the voice was made with permission.
- Clients, platforms and stores increasingly ask for that proof.
- Clear limits prevent disputes later.
Record a clean sample first
Before anyone signs, let the speaker hear what a synthetic voice sounds like: on the page of any catalog voice, type a line of 30 to 100 characters and press Voice it. Understanding the result makes consent informed.
What the form should cover
Add a line on disclosure: whether listeners will be told the voice is synthetic, and whether the speaker’s name may be credited.
- Who: the speaker’s name and contact, and who receives the right to use the voice.
- What: which recordings are used and that a synthetic voice model will be trained from them.
- Uses: allowed purposes, such as narration for your courses or a game character, and forbidden ones, such as political messages or impersonation.
- Terms: duration, territory, payment, how the speaker can withdraw consent, and how the recordings and the voice file are stored and deleted.
Using the voice in SuperClonica
In the SuperClonica studio, a voice is trained from a clean recording of 5 to 10 seconds on a computer with an NVIDIA graphics card, or ordered on the exchange from an author who trains it. Keep the signed consent with the recording and the resulting voice file. The Supertonic 3 model license, OpenRAIL-M, forbids using speech to impersonate someone without consent, to deceive or to harass, and the SuperClonica terms forbid presenting synthetic audio as a real person’s words.
Good practice after signing
Store the recordings and the voice file securely and only as long as the consent allows. Keep a simple log of where the voice is used. If the speaker withdraws consent, stop new uses and delete the files as the form says. Review the form when you start a new kind of project with the same voice.