← Blog

I made a voice in ElevenLabs. Can I register it?

2026-11-04 · OpenStela

Yes, you can register what you made with the voice: the narration, the audiobook chapter, the vocal take, the video with the voiceover. Each of those is a file, and a file has a fingerprint and a date. What you can't register is the voice itself. A voice isn't a file. And under US law it isn't something copyright covers in the first place.

That difference matters more for voice than for almost any other AI output. If the voice sounds like a real person, the rules that apply are about that person's identity, not about copyright, and a dated record of your file won't help you there. This post goes through what you actually hold, how to register it, and where the voice-rights risk sits.

A voice isn't a work. A recording is.

The US Copyright Office put this plainly in Part 1 of its report on copyright and AI, Digital Replicas, published July 31, 2024. On page 17: "Copyright does not, however, protect an individual's identity in itself, even when incorporated into a work of authorship. A replica of their image or voice alone would not constitute copyright infringement." The same page quotes the Ninth Circuit in Midler v. Ford Motor Co. (1988): "A voice is not copyrightable. The sounds are not 'fixed.'" (Part 1 report, PDF; report index)

A recording is a different thing. The Office's Circular 56 (revised March 2021) quotes the Copyright Act's definition of sound recordings as "works that result from the fixation of a series of musical, spoken, or other sounds." Spoken is in the list. So a narration can be the kind of thing copyright deals with, even though the voice in it isn't.

Whether your narration is protected is a further question. When the sound was produced by an AI model, the Office's position in Part 2 of the same report (published January 29, 2025) is that protection depends on the human contribution, not on the prompt alone. Can AI-generated art or music be copyrighted? goes through that with sources. Your script, your edit, your mix and the choices you made are where your own contribution is clearest.

This is general information, not legal advice.

What you can actually register

The last row needs a caveat. ElevenLabs isn't one of the sites we read specifically. A link to it is handled like any other web page: we fetch it at that moment, save the snapshot, and fingerprint the snapshot. If the page can't be fetched, you get an error and nothing is registered. We haven't tested what an ElevenLabs page returns to a plain request, so register the exported file if you have it.

Whose voice is it? Three cases

Before you register anything, be honest with yourself about which of these you're in.

A voice you designed from scratch. You described a voice or tuned one, and it isn't modelled on anyone. This is the simplest case. The open question is the ordinary AI one: how much of the output is yours.

Your own voice, cloned. It's your identity, so the identity question mostly goes away. You still need to follow whatever ElevenLabs requires for cloning, and the audio you produce is still AI output, with the same contribution question as above.

Someone else's voice. A friend, a voice actor, a celebrity, a character's original performer. This is where the risk is, and it isn't copyright risk. Many US states protect a person's identity, including their voice, through what's called the right of publicity. The Copyright Office's Part 1 report describes California Civil Code § 3344 as "protecting against the unconsented commercial use of a person's name, voice, signature, photograph, or likeness" (page 12), and notes that Tennessee "recently extended its right of publicity statute to encompass voice simulations" (page 15). Other states differ, and the report spends several pages on how much.

In Taiwan, the Civil Code's Article 18 says that when a personality right is infringed, you can ask the court to remove the infringement, or to prevent it when it's threatened (民法 第 18 條, current text last amended 17 August 2026). The article doesn't list voice by name, and we didn't find an official Taiwanese source on voice specifically, so we won't guess how a court would read it.

If the voice belongs to someone else, get their written permission before you publish anything. A record with us won't change that. A dated record of a file that clones someone without consent is just a dated record that you published it.

How to register the audio

1. Pick the final export, the file you're actually going to publish. A re-export, a format conversion or a loudness pass gives a new file with a new fingerprint, so register the one you'll release and keep it.
2. Open the works page, choose Register a work, then Upload a file. Audio can be mp3, m4a or wav, up to 30 MB per file. Video can be mp4 or webm, up to 95 MB.
3. Attribution: the name you publish under. A pen name is fine. Created on: the date you made it, or leave it blank if you're not sure.
4. Tick This work was made with AI tools and write "ElevenLabs" under Which tools, plus whatever else you used: your editor, your DAW. Search and feeds on OpenStela can be filtered by tool, and ElevenLabs is one of the filters, so this is also how people find it.
5. Write Your contribution. It's required for AI-assisted work, and it goes into the record. Be specific: "Wrote the script; designed the voice from a text description; generated each paragraph several times and kept the best takes; cut and mixed in Audacity; added music I composed." Say which kind of voice it is. If it's your own voice cloned, say so.
6. Rights: choose My own work or I hold a licence. For any audio file, the form also asks you to confirm two things separately, the composition and the recording. It was built with songs in mind, and there's no separate option for speech. Only tick what's true for you. If you can't honestly tick both, don't register that file.
7. Decide whether it shows on your public page. If it isn't out yet, leave it private. The record is anchored in that night's batch either way, and you can make it public later.

Registration is free. Each account has 300 MB of space for uploaded works, and there's a daily cap on how many works one account can register (see pricing). You get a public page for the work, a downloadable record and a Proof Pack that can be checked without us. Once a song is registered, how do you prove it existed that day walks through the check; the same steps apply to spoken audio.

If the narration belongs to a character, say an AI host or a virtual narrator, you can bind it to that character too. That's a second dated record, covered in how to prove a song belongs to a character that has no face. The same steps work for spoken audio.

Read ElevenLabs' terms for your plan yourself

What you may do with ElevenLabs output, including commercial use, attribution and which voices you may clone, is set by ElevenLabs, and it can depend on your plan. When we checked on 4 November 2026, the ElevenLabs Terms of Service (non-EEA) and the Prohibited Use Policy both showed "Last Updated: 09 October 2026". The page we read is the non-EEA version, so if you're in the EEA, check which version applies to you.

We won't paraphrase those pages. They change, and a summary goes stale before the page does. Registering a file with us doesn't change what ElevenLabs allows. If their terms don't let you use a voice the way you plan to, a dated record doesn't make it all right.

What the record shows, and what it doesn't

A record shows that a specific file existed on a specific date, under the name you gave, with the tools and contribution you wrote down. If the same narration turns up under someone else's name later, you have a dated, verifiable record to point to. That's evidence if a dispute comes up, and nothing more. What a character ID proves, and what it doesn't has the longer version.

It doesn't say whose voice is in the file. We don't listen to the audio, compare it with other recordings, or check your contribution text. It doesn't make you the owner of a voice, it doesn't settle whether the recording is protected, and it doesn't stop anyone from cloning a voice that sounds like yours. If someone does copy your work, Someone copied my AI character. What can I actually show? covers the next step.