Real consent for AI voice cloning means the person whose voice it is reviews what will be built, accepts a specific statement, and leaves a dated record, all before any audio reaches a cloning model. A checkbox buried in signup terms does not do that.
Jellypod's voice cloning works this way. We save a versioned, timestamped consent record first, and only then send samples to the voice provider. A guest can consent for themselves through a single-use invite link, but nobody can consent for them.
Key takeaways
- Consent has to come from the voice's owner, be specific to cloning, and be recorded before samples reach the provider.
- Ask any tool who submits the audio, what text was accepted, when the record is written, and who can edit it afterward.
- Jellypod stores the exact attestation text, its version, a timestamp, the accepting user or invite, and the request's IP address and user agent, tied to the resulting voice.
- Guests clone their own voice through a single-use invite link that expires after 24 hours, with no Jellypod account needed.
- A record proves an attestation was accepted. It does not prove identity, so keep your own permission in writing for anyone you invite.
Jellypod writes the consent record at step 3, before any audio reaches the voice provider.
What does consent mean for AI voice cloning?
Consent means the person heard, in plain terms, that their voice would be cloned, and said yes to that specific act. Agreeing to a platform's general terms during account creation does not cover it.
Consent is also different from licensing. Consent answers whether a voice can be cloned at all. A license answers how long, for which projects, and whether the clone can be reused later. A tool that stores only a yes-or-no flag cannot represent either. Where you need licensing terms, for example with a hired voice actor, put them in a written agreement outside the tool.
What should a consent-first voice cloning tool require?
Five things, in this order:
- The owner submits the audio. Not a producer uploading an old interview on their behalf. A recording being public is not permission to clone it.
- A specific statement is accepted. The person confirms the samples are their voice and that they hold the rights to clone it.
- The acceptance becomes a record. The exact text shown, its version, a timestamp, and who accepted it, not a stored boolean.
- The record exists before the provider gets audio. If the approval happens in parallel or afterward, a failure partway through leaves a voice that was never authorized.
- Users cannot rewrite the record. Consent history should not be editable by the people it protects or the people it binds.
Ask a vendor which of these it can show you.
How does Jellypod handle voice cloning consent?
Every clone path in Jellypod writes a consent record before calling the voice provider. There are two paths.
Cloning your own voice in the Studio. On Characters, you choose Create Voice Clone, record or upload at least 30 seconds of audio (60 seconds is recommended), and tick an attestation on the same step: "I confirm that these samples contain my voice and that I have all rights and consents required to upload and clone it. I will not use the cloned voice for illegal, fraudulent, deceptive, or harmful purposes." You also agree to the Terms of Service and Privacy Policy. Jellypod then saves the consent record, and only after that does it clone the voice and transcribe the samples. You review the drafted name, language, and backstory afterward. The steps are in the voice cloning help doc.
Cloning a guest's voice. You cannot upload a guest's audio and attest for them. You send an invite link instead, and the guest opens it, records or uploads their own samples, and ticks the same attestation themselves. The link is single-use and expires after 24 hours, and the guest needs no account. We wrote up the flow in the voice clone invites announcement. Once the guest submits, their voice becomes a host in your library.
The server ignores any policy text the browser sends and writes its own. Each record stores:
- the exact attestation text and its version (currently 2026-08-24.1)
- links to the Terms of Service and Privacy Policy in force
- the acceptance timestamp
- the accepting user, or the invite ID for a guest
- the subject's name and whether they are the account holder or an invited guest
- the request's IP address and user agent
The record is linked one-to-one to the resulting voice. A database trigger rejects updates, so a record cannot be edited after insert, and ordinary users cannot read or change these records. Deletion happens only when an organization is torn down. The audio samples are not kept as evidence, and if the provider call fails, the record of that attempt remains.
What does a consent record not prove?
It proves that someone clicked an attestation from a given session at a given time. It does not prove the person is who they claim to be, and Jellypod does not run a voice-match check against the samples. A stranger could submit someone else's audio and lie in the attestation.
So an invite is a good control for people you know, and a poor one for people you do not. If you are cloning an expert, a speaker, or a colleague, get their permission in writing as well, and say what the clone will be used for. That is the practical difference between consent and licensing, and the tool cannot give you the second one.
Jellypod's plans also cap how many clones an account can hold: 2 on Starter, 8 on Creator, and 20 on Business.
Is it legal to clone someone's voice without consent?
It depends on where you are, and the rules are changing. Here is what we could verify from primary reporting, current as of September 2026.
| Jurisdiction | Status | What it would do | Source |
|---|---|---|---|
| United States | NO FAKES Act (S. 4591) advanced out of the Senate Judiciary Committee by voice vote on June 18, 2026, not yet law | A federal, licensable property right in a person's own voice and visual likeness, plus a notice-and-takedown process | Holland & Knight |
| Denmark | Copyright amendment unveiled June 26, 2025, with enactment forecast for late 2025 or early 2026 (we have not confirmed it passed) | Rights over a person's voice, facial features, and body, including the right to have unauthorized digital reproductions removed | Schjodt |
Until a federal law passes, US protection comes from state right-of-publicity statutes, which vary. This is not legal advice. If your project involves a paid performer or a public figure, ask a lawyer.
Can you clone your own voice for a podcast or video?
Yes. You are both the subject and the person accepting the terms, so the third-party questions above do not apply. Record 60 seconds of clean audio in a quiet room, and you can use the clone for a podcast host, a narrator, or the voice of a faceless YouTube channel.
Cloned voices still need direction to sound natural. Our guide to making an AI voice sound human covers pacing, script editing, and pronunciation. If you want a voice you do not have to clone, browse the AI voice library, or see how to change the voice in a NotebookLM podcast.
Frequently asked questions
Do I need consent to clone a voice I found online?
Yes. A public recording does not grant permission to clone the voice in it. Ask the speaker directly, and use a tool that has them submit their own samples.
Can a guest take their clone back?
Consent records are retained for legal and takedown work, and the takedown process itself is still being built. Agree with your guest on how the clone will be used and retired before you send the invite.
How long does a guest have to accept an invite?
24 hours from creation. You can resend or revoke pending invites from the Hosts page in your account.
