ElevenAI: ElevenLabs' Free AI Teammate for Meetings, Explained

ElevenLabs spent three days in November shipping like a company trying to outrun its own reputation. Campaign Manager landed on November 9. Scribe v3 and Scribe v3 Live followed on November 10. Then, at its New York Summit on November 11, the company unveiled 13 new products and upgrades, and the one aimed at ordinary workers rather than developers or contact centers was ElevenAI: an app that listens to your meetings and hands back the notes.

Free early access opened the same day. ElevenLabs has not published a price for what comes after it, has not said when early access ends, and has not listed which platforms the app runs on. That combination is the whole story of ElevenAI right now. It is a real, usable app built on an improved transcription model, and it is also a free trial with no visible end date and no conversion terms. If you already pay for a meeting notetaker, the smart move is to run ElevenAI next to it for a fortnight before your next renewal, not to switch on day one.

What Is ElevenAI?

ElevenAI is a new workplace app from ElevenLabs, introduced at the company's New York Summit on November 11, 2026. ElevenLabs describes it as an AI teammate for enterprise teams. It listens to spoken conversations, including in-person meetings, and turns them into notes, summaries and action items. The company frames it as the employee-facing counterpart to ElevenAgents, its platform for customer-facing voice and chat agents. Where ElevenAgents handles the calls your customers make to you, ElevenAI is meant to handle the meetings your own team sits in.

Three things happen after a meeting ends. The app delivers a summary and a list of action items to each attendee. It also offers dictation into any app, so the same voice engine that takes the notes can type where you point it. And it can give live assistance during a call, sharing context between people and agents while the conversation is still running. ElevenLabs says the app adapts to a company's tools and processes and passes context between agents and people so they can work together as one team.

That last part is doing a lot of work. Most meeting notetakers stop at the transcript. ElevenAI is pitched as something that stays with people through the workday, which is a bigger promise and, for now, a mostly unproven one. The meeting notes are live today. The wider vision of an app that follows your work across tools is a direction, not a delivered feature set.

What ElevenAI Actually Does Today

Strip away the keynote language and ElevenAI does four concrete things, all confirmed in ElevenLabs' own Summit write-up and in independent coverage of the launch.

Meeting notes with consent built in. The app takes notes in meetings only with everyone's consent, according to ElevenLabs. Notice and consent mechanisms are built in rather than added later, which matters more than it sounds. An app that listens to in-person meetings has a legal profile that changes by country and, in the United States, by state. ElevenLabs building the notice step into the product is the right call, but the compliance obligation still sits with you. If you record a meeting in a place where all-party consent is required, the app's built-in notice does not replace your responsibility to get it.

Per-attendee summaries and action items. When a call ends, each attendee gets a summary and action items. That is a small detail with real consequences. Most notetakers produce one shared document and leave it to someone to split the work. ElevenAI assigns the follow-ups per person, which is the part of meeting notes that usually never happens.

Dictation into any app. ElevenAI does dictation, which puts it in the territory of tools like Wispr Flow as well as meeting notetakers like Fireflies.ai. You speak, it types, in whatever app you are working in. For anyone who writes long emails, documents or messages by voice, this is the feature that could justify keeping the app open all day rather than only during meetings.

Live assistance during calls. The app can give live assistance while a call is running and pass context between people and agents. ElevenLabs has not published detail on what that assistance looks like in practice, which answers it can give, or which tools it can reach into. Treat it as a live feature with thin documentation until the company says more.

Scribe v3: The Engine Underneath

ElevenAI's dictation and meeting notes run on Scribe v3, the speech-to-text model ElevenLabs launched one day earlier, on November 10, 2026. The model is the reason ElevenAI is worth taking seriously even though the app itself is brand new.

ElevenLabs says Scribe v3 makes 27% fewer word errors than Scribe v2 across 102 languages. That is the company's own benchmark, against its own previous model, on its own test audio. It is a real improvement and a checkable claim, but it is not an independent comparison against competitors. Scribe v3 is built for recorded audio such as meetings and voice notes, and for messy real-world audio: heavy background noise, and telling a pause apart from the end of a sentence.

Scribe v3 Live is the separate real-time variant, and it is arguably the bigger news of the two. It supports speaker diarization for the first time in a live ElevenLabs model, and ElevenLabs says it makes 43% fewer semantic errors than its predecessor, with turn detection that reads the whole conversation rather than the last moment of audio. Scribe v3 Live is already in ElevenAgents, and both models are in ElevenAPI. Scribe v3 is now the default transcription model in the ElevenLabs app.

One feature the launch video shows that the blog post never mentions is custom formatting rules. You can tell Scribe v3 to write all dates as YYYY-MM-DD and it applies that across the transcript, and there is a toggle to render numbers as digits or words. For anyone producing transcripts to a house style, that is a bigger practical win than another percentage point of accuracy.

There is also a healthcare-tuned variant, Scribe v3 Medical, which recognises thousands of clinical terms. ElevenLabs' own example is telling cycloserine apart from cyclosporine, a tuberculosis drug and an immunosuppressant. A financial-services domain model is described only as in the works.

What It Costs: Free Now, No Price After

Here is the honest pricing picture, because it is the part of this launch that most coverage glosses over. Free early access to ElevenAI opened on November 11, 2026. ElevenLabs has published no price for what comes afterwards, no end date for the free period, and no terms for converting early-access users to a paid plan. Free early access is a recruitment offer, not a price. It tells you nothing about what the paid tier will cost when early access closes, and ElevenLabs has not said.

What you can price today is the engine underneath it. Scribe v3 costs $0.20 per hour of audio through the ElevenLabs API, and Scribe v3 Live costs $0.22 per hour, according to ElevenLabs' own pricing page. Those are the same rates as the Scribe v2 models they replace, so the accuracy upgrade comes with no price increase. At November 13, 2026 market rates (1 USD = 0.7544 GBP, 1 USD = 1.4249 CAD), that is about £0.15 or CA$0.28 per hour for Scribe v3, and about £0.17 or CA$0.31 per hour for Scribe v3 Live. Keyterm prompting, which lets you feed the model up to 1,000 custom terms such as product names, adds $0.05 per hour on Scribe v3 and $0.08 on Scribe v3 Live.

Put that in meeting terms. A team that records 20 hours of meetings a month would pay about $4.00 a month for Scribe v3 transcription through the API, before any plan or platform fee. That is the raw transcription cost, not the price of ElevenAI itself, which remains unpublished.

ElevenLabs' own subscription plans give a sense of the floor. The free plan costs $0 and includes 10,000 credits a month. Paid plans start at $6 a month for Starter, according to pricing coverage checked in early November 2026. Speech-to-text costs 330 credits per minute of audio on the credit system, which means the free plan's monthly credits cover roughly 30 minutes of transcription a month. That is enough to test ElevenAI on a few real meetings, and not enough to run a team on it. Two more limits apply to the free plan: content made on it cannot be used commercially, and anything you publish from it must credit ElevenLabs.

One more cost warning. Scribe v3 and Scribe v3 Live run in the United States by default. ElevenLabs offers isolated environments in the EU, India and Singapore, but only to Enterprise customers. If your team has data-residency requirements, the $0.20 hourly rate is not the price you will actually pay.

ElevenAI vs Fireflies: The Honest Comparison

ElevenAI does dictation and meeting notes, which is squarely the territory of Fireflies.ai, the most established independent meeting notetaker. Here is how the two line up on the facts available today.

The table shows why ElevenAI is interesting and why it is not yet a switch. Fireflies is a finished product with published prices, published limits and years of integrations. ElevenAI is free, covers more languages, and does dictation that Fireflies charges separately for, but it has no published price, no published limits and no stated platform list. If you run five meetings a week, Fireflies Pro costs $120 a seat a year on annual billing. ElevenAI costs nothing today and an unknown amount tomorrow. That unknown is the decision, not the feature list.

For the wider field, our roundup of the 7 best AI meeting note-takers in 2026 compares bot-based and botless options across the same questions: price, limits and what each plan actually buys. ElevenAI would sit in the botless camp, since it listens rather than joining as a visible bot, but it is too new for a ranked slot until its pricing lands.

What Is Still Missing

About half of what ElevenLabs announced across November 9 to 11 is preview, early access or roadmap rather than generally available. The keynote does not draw that line for you, so here it is.

Interaction Engine is a preview, not a release. It manages conversational rhythm, turn-taking and interruptions, and routes each part of an exchange to an appropriate model: a cheaper model for simple requests, a frontier model for complex ones like refunds. If it works, it is a direct cost lever on every agent deployment. No date has been given.

Memory and Routines are coming, not here. ElevenAgents will add Memory, which builds a record of a returning customer over time, and Routines, which automates multi-step work before, during and after conversations. Examples include pre-screening a mortgage applicant and calling back within minutes, or reminding patients before a visit and offering a cancelled slot to a waitlist. Both are previews with no dates. They are also enterprise features in an enterprise product, so do not expect either in a Starter plan.

Custom model fine-tuning starts in the coming months. Enterprises will be able to bring their own data and fine-tune text to speech, speech to text and open-source language models, picking their own target metrics and working with ElevenLabs researchers. The CEO compared it to each company having its own reinforcement-learning stack. That is the announcement to come back for, and it is also the one with the least detail today.

Platforms, retention and data handling are unstated. ElevenLabs calls ElevenAI an app but has not listed iOS, Android, web or desktop availability. The Summit post gives no retention policy, no statement on how long recordings are kept, and no detail on what happens to your notes if you later stop paying. Those are the questions to ask before putting client meetings into it.

No independent accuracy tests exist yet. The 27% and 43% figures are ElevenLabs' own benchmarks. Scribe v3 only shipped on November 10, so no independent lab has published a comparison. Treat the accuracy claims as promising and unverified by third parties.

One more flag from the Summit package. ElevenLabs says its expanded NVIDIA partnership puts Nemotron 3.5 Super VL into ElevenAgents, and that the model completed its customer-support benchmark tasks at 70% lower cost than other leading open models. That is ElevenLabs' own benchmark, not independently verified.

How to Try ElevenAI

Getting started takes minutes, and the free early access means there is no payment step to clear first.

First, create a free ElevenLabs account. ElevenAI lives inside the ElevenLabs ecosystem, so you need an account before the app appears. Second, join the ElevenAI early access. Because free early access opened on November 11, 2026, ElevenAI is listed as an early-access app, so it may appear with an early access or try-now label rather than as a finished product. Third, run it on one low-stakes internal meeting, not a client call, and check the transcript against what was actually said. Scribe v3's custom formatting rules are worth setting up at this stage: tell it how you want dates and numbers written, and it applies that across the transcript. Fourth, compare the per-attendee action items against your own notes. If the app misses owners or deadlines, you have learned that before it matters.

For context on where voice agents are going, our review of Mistral's Voxtral 3 covers a phone-ready voice agent at $0.04 a minute, a useful price anchor for what live voice actually costs once you leave the free tier.

Common Mistakes to Avoid

The first mistake is treating free early access as a price. It is not. There is no published end date and no published price for what comes next, so build your workflow on the assumption that a bill arrives, and keep an exit plan for your notes. The second mistake is recording a meeting without telling everyone in the room. ElevenAI builds notice and consent mechanisms in, but the legal duty is yours, and it varies by jurisdiction. Get consent in the open, every time. The third mistake is judging Scribe v3 on one clean call. Its claims are about messy audio, background noise and overlapping speech, so test it on your worst meeting, not your best. The fourth is assuming the free plan covers a team. Ten thousand credits a month is roughly 30 minutes of transcription, which is a trial, not a rollout. The fifth is buying the roadmap. Interaction Engine, Memory, Routines and fine-tuning are previews. If you need them, wait until they ship.

Frequently Asked Questions

Is ElevenAI really free?

Yes, for now. Free early access opened on November 11, 2026, and you can start without entering payment details. ElevenLabs has not published pricing for what comes afterwards, and there is no stated end date for the free period, so treat the free period as a chance to learn the tool, not a promise that it will stay free forever.

What is the difference between Scribe v3 and Scribe v3 Live?

Scribe v3 is the batch transcription model for recorded audio, and ElevenLabs says it makes 27% fewer word errors than Scribe v2 across 102 languages. Scribe v3 Live is the real-time variant for voice agents and live captions. It adds speaker diarization for the first time in a live ElevenLabs model and, by the company's own testing, makes 43% fewer semantic errors than the previous live model. Through the API, Scribe v3 costs $0.20 per hour of audio and Scribe v3 Live costs $0.22.

Does ElevenAI record in-person meetings?

Yes. ElevenLabs says ElevenAI listens to spoken conversations including in-person meetings, and takes notes with everyone's consent. Notice and consent mechanisms are built into the app. The company has not published detail on recording retention or data handling, so check those terms before using it for sensitive meetings.

What happens to my notes when early access ends?

ElevenLabs has not said. There is no published end date for free early access, no conversion terms, and no statement on what happens to stored notes if you stop paying. Export anything important while access is free, and do not treat the app as your only copy of meeting records.

Does ElevenAI work with Zoom and Google Meet?

ElevenLabs has not published a platform or integration list for ElevenAI. The Summit announcement describes meetings, calls and in-person conversations without naming specific platforms. Until the company lists integrations, assume support is unconfirmed for any particular meeting tool and test it on your own setup first.

How accurate is Scribe v3, really?

ElevenLabs says Scribe v3 makes 27% fewer word errors than Scribe v2, and Scribe v3 Live makes 43% fewer semantic errors than its predecessor. Both figures are the company's own benchmarks on its own test audio, not independent comparisons. No third-party lab has published Scribe v3 accuracy results yet, so the honest answer is that the claims are promising and untested by outsiders.

ElevenAI is the most useful thing ElevenLabs shipped in its three-day run, because it is the only one you can open today for free and use on your next meeting. The transcription engine underneath it is measurably better than the one it replaces, and it costs the same per hour. What is missing is everything around the free offer: a price, an end date, a platform list and a retention policy. Use it now, on real meetings, next to the notetaker you already pay for. Let the comparison, not the keynote, decide whether it earns a subscription when the free period ends.

Sources

• ElevenLabs, "Building the interaction layer for AI: Everything announced at ElevenLabs Summit NYC," November 11, 2026 (official announcement)
• ElevenLabs pricing page (Scribe v3 $0.20/hr, Scribe v3 Live $0.22/hr; free plan 10,000 credits/month)
• Unite.AI, "ElevenLabs Unveils Model Customization, ElevenAI and Partner Program," November 12, 2026
• DailyAIPedia, "ElevenAI Free Access, and What's Next From ElevenLabs," November 12, 2026
• DataNorth, "ElevenLabs Scribe v3: 27% fewer errors, same price," November 11, 2026
• CellCog, "ElevenLabs Scribe v3: Accuracy Claims, Prices, First Test," November 12, 2026
• Voibe, "Fireflies Pricing: Pro Is $10 a Month. Is There a Free Plan?" (pricing checked October 26, 2026)
• Fireflies.ai pricing page (Free $0, Pro $10/seat/month billed annually)