ElevenLabs and the AI voice business

ElevenLabs began in 2022 with a text-to-speech model. Four years on, it sells synthetic speech, voice cloning, dubbing, music, sound effects and conversational agents, and it says private investors now value it at $22 billion.[1][2] Its tools also sit at the centre of the debate about cloned voices and consent.

This profile sets out what the company makes and charges, as its own pages describe them, and what a buyer has to weigh. It is not a review: AI Chronicle has not tested ElevenLabs products, and we make no quality claims. For the production and editing tools themselves, as ElevenLabs documents them, see our ElevenLabs review. Figures the company reports are attributed to it. All pages were read on 16 October 2026. For the wider picture of generated media, start with our explainer on how AI video generators work.

What ElevenLabs makes

The company describes two platforms built on one research foundation, ElevenCreative and ElevenAgents, plus ElevenAPI for developers.[1]

• ElevenCreative generates speech, music, sound effects, images and video. It includes voice cloning, voice design from a prompt, a voice library and an editor for audiobooks and voiceovers.[1]
• Dubbing translates audio and video into other languages. The documentation says it aims to preserve "the unique characteristics of the original speakers' voices".[3]
• ElevenAgents lets organisations configure, deploy and monitor agents that work across phone, chat, email and WhatsApp.[1]
• ElevenAPI exposes the models directly: text to speech, transcription, music, sound effects, voice changing and dubbing.[4]

The company says its music model was "built in partnership with artists, labels, and publishers".[1] It also runs an Iconic Marketplace, where companies ask rights holders to license the voices of figures such as Sir Michael Caine. Each request is approved or refused by the talent or rights holder.[5]

AI Chronicle

How it charges

The three lines are priced on different units, so plans are not directly comparable.[6][7][4]

Rights depend on the tier. A commercial licence and Instant Voice Cloning start on the $6 Starter plan, and Professional Voice Cloning starts on the $22 Creator plan.[6] The prohibited-use policy bars commercial use on the free plan.[8] The pricing pages were showing first-month discounts on the day we checked, and prices exclude taxes.[6]

Funding and scale, as the company reports them

On 4 February 2026, ElevenLabs announced a $500 million Series D led by Sequoia Capital, at an $11 billion valuation. It said the round brought total funding to $781 million across five rounds, and that it closed 2025 with over $330 million in annual recurring revenue.[9]

On 30 September 2026, it said a $300 million employee tender offer, led by Wellington and T. Rowe Price, valued it at $22 billion.[2] A tender lets employees sell existing shares to investors, so it is not new capital for the company. TechCrunch reported the tender and the doubling from the February valuation.[10] In the same announcement, the company said enterprise accounts for "55% of our revenue". It also said its agents handle more than 15 million conversations a week and that it has over 800 staff.[2] We have not independently verified these figures.

The safeguards it describes

The company's safety page groups its measures into four layers.[11]

• Prevent: red-teaming models before release, vetting customers at sign-up, blocking the cloning of "celebrity and other high risk voices", and requiring technological verification for Professional Voice Cloning.
• Detect: AI classifiers, human reviewers, internal investigations and a public abuse-report form.
• Enforce: bans for persistent or serious violators, and referral of criminal activity to law enforcement.
• Inform: use of the C2PA provenance standard and a public AI Speech Classifier.

The documentation turns some of these into product rules. A Professional Voice Clone must pass a voice check:

“For now, we only allow you to clone your own voice.”

— ElevenLabs documentation, Professional Voice Cloning.[12]

Instant clones work differently. The user confirms they have "the right and consent to clone the voice".[13] That is a declaration, not a voice check.

The prohibited-use policy forbids cloning a living person's voice without "explicit and informed consent or other valid authorization". It also forbids impersonating candidates without authorisation and circumventing verification systems such as "Voice CAPTCHA".[8]

The AI Speech Classifier estimates whether a clip was made with ElevenLabs. Its own page notes the limits: it "does not reliably classify audio generated with the Eleven v3 model" and cannot detect other providers' audio.[14]

ElevenLabs is listed as a general member of the C2PA.[15] The company says plainly that its layers can fail:

“We recognize that no safety system is perfect: on occasion, safeguards may mistakenly block good actors or fail to catch malicious ones.”

— ElevenLabs, Safety.[11]

What to weigh

Consent is yours to get and keep. Platform checks govern what you can do on the platform. They don't give you rights in someone's voice. Our AI voiceover workflow sets out a consent record and review gates.

Likeness law is spreading. Tennessee's ELVIS Act took effect on 1 July 2024 and added voice to the state's protected personal rights.[16][17] It creates civil liability for unauthorised public use of a person's voice. It also covers distributing a tool whose primary purpose is producing "a particular, identifiable individual's photograph, voice, or likeness" with knowledge that this was not authorised.[17] The US Federal Trade Commission's rule on impersonating government agencies and businesses took effect on 1 April 2024. The FTC also proposed covering the impersonation of individuals and the provision of "means and instruments" to scammers.[18] Voice rights are a separate question from copyright in training data, which our analysis of AI training and copyright covers. This is not legal advice.

Disclosure follows the use. The policy requires users to disclose AI where the law requires it or where silence "would materially mislead".[8] For agents answering the public, that decision comes before launch. Our profile of Vapi's approach to voice agents looks at the same question from the infrastructure side.

Provenance helps only if it survives. Signed metadata can be stripped. See our explainer on content credentials.

Hypothetical: one company, four uses

This is a hypothetical example. It describes no real customer.

A training company wants four things. The product lines and rules below come from the published pages; the "what it must still do" column is our editorial suggestion.

Questions readers commonly ask

Is ElevenLabs only a text-to-speech company?

No. It still sells text to speech, but it also sells transcription, dubbing, music, sound effects and an agents platform. It names ElevenAgents as its fastest-growing product.[2]

Can I clone a client's voice for them?

Not as a Professional Voice Clone. The voice owner must create and verify it in their own account, then share it.[12] For any clone, you need their explicit consent.[8]

Can the classifier tell me whether any clip is AI?

No. It checks only for ElevenLabs audio, and its own page says it does not reliably flag one of the company's models.[14]

Did ElevenLabs raise $300 million in September?

No. That was an employee tender at a $22 billion valuation. Its last primary round we found was the $500 million Series D in February 2026.[2][9][10]

The next step

If you are about to produce narration with any synthetic voice, follow our guide to building an AI voiceover workflow with review.

About these sources

All pages were accessed on 16 October 2026. Company pages establish what ElevenLabs says it offers, charges or has achieved; they do not show how well its products or safeguards work. AI Chronicle has not tested ElevenLabs products. The figure and the hypothetical table are illustrative.

Related reading

• How to keep characters consistent in AI video
• Vapi's approach to voice agents