How ElevenLabs Grew to $600M ARR

ElevenLabs grew by making AI voices that sounded much better than what came before, letting anyone try them, and turning the audio people created into its distribution.

The company did not start with the product that eventually took off.

Its founders first wanted to build AI dubbing. They pitched the idea to thousands of YouTubers, but the response was weak. What creators wanted was simpler: better narration, voiceovers, and a way to generate speech without hiring a voice actor.

ElevenLabs followed that demand.

It launched a text-to-speech beta in January 2023. Five months later, it had more than 1 million registered users.

From there, ElevenLabs expanded into APIs, multilingual speech, dubbing, a marketplace for licensed voices, and finally voice agents for large companies.

The business grew just as quickly.

ElevenLabs said it ended 2025 at roughly $350 million in annual recurring revenue. It passed $500 million ARR in the first four months of 2026. By July, CEO Mati Staniszewski said the company had reached around $600 million ARR.

These are company-reported annualized figures, not audited revenue.

The interesting part is not just how fast ElevenLabs grew.

It is how one growth loop led to the next.

Creators spread the product. Developers embedded it. Voice actors supplied the marketplace. Then enterprises turned the technology into customer support, sales, and other workflows.

ElevenLabs growth timeline

Date	Milestone
2022	ElevenLabs is founded by Mati Staniszewski and Piotr Dąbkowski
Jan 2023	Public beta launches
Jun 2023	Passes 1 million registered users
Aug 2023	Launches multilingual speech across 29 languages
Oct 2023	Launches AI dubbing
Feb 2024	Adds payouts for voice actors
Late 2024	Starts expanding into conversational voice agents
2025	Startup Grants passes 4,000 startups
End of 2025	Reaches roughly $350M ARR
Feb 2026	Raises $500M at an $11B valuation
May 2026	Passes $500M ARR
Jul 2026	CEO says ARR has reached roughly $600M
Sep 2026	Large organizations account for 55% of revenue

The timeline makes ElevenLabs look like a company that knew exactly where it was going.

The real story was messier.

ElevenLabs started with dubbing. Creators wanted something else.

ElevenLabs was built around a problem its founders knew well.

Mati Staniszewski and Piotr Dąbkowski grew up in Poland, where foreign films were often translated with poor dubbing.

They thought AI could translate a movie into another language while keeping the original actor's voice and emotion.

It was a big idea.

But before going all in, Staniszewski tried to find out whether creators actually wanted it.

He sent personalized messages to thousands of YouTubers.

He has said the first batches got roughly a 15% reply rate, but the conversations were not promising. Creators doubted the technology, could not see how dubbing would fit their workflow, and YouTube did not yet have strong support for multiple audio tracks.

So the founders narrowed the problem.

Instead of trying to solve dubbing all at once, they focused on one part of it:

making generated speech sound human.

That decision changed the company.

When ElevenLabs shifted from dubbing toward narration and voiceovers, it started to see a much stronger product signal.

This is an important part of the ElevenLabs story.

The founders did not get product-market fit from the product they originally wanted to build.

They found it by paying attention to the smaller jobs people actually wanted done.

An author showed them what users really wanted

One of the clearest signals came during ElevenLabs' closed beta.

The early product had a small text box. It was meant for generating short pieces of speech.

An author used it to make an entire audiobook.

They pasted the book into the product piece by piece, generated the audio, downloaded hundreds of files, and stitched everything together manually.

It was a terrible workflow.

That was exactly why it mattered.

Most people will not fight a product for hours unless the result is valuable to them.

Staniszewski has described this as one of the first signs that ElevenLabs had found something people really wanted.

The user had discovered a use case before ElevenLabs had properly built the product for it.

The lesson was simple:

watch what users are willing to suffer through.

A feature request tells you what someone says they want.

A user manually repeating the same action hundreds of times tells you what they actually want.

For ElevenLabs, that demand was showing up around narration, audiobooks, video voiceovers, games, and other forms of generated speech.

A better voice model created the first growth loop

ElevenLabs still needed people to notice the product.

The quality of the model did most of that work.

Before ElevenLabs, text-to-speech was already common. Google, Amazon, Microsoft, and others had offered it for years.

But synthetic speech still often sounded synthetic.

ElevenLabs focused heavily on emotion, pacing, tone, and context.

In January 2023, the company published an early demonstration called "This Voice Doesn't Exist".

One of the things that caught people's attention was that the generated voice could laugh.

The demo spread through technical communities and newsletters.

A Hacker News discussion received hundreds of comments, with the founders answering questions about voice cloning, latency, and API access.

Staniszewski later said the launch brought roughly 1,000 people onto the waitlist the next day.

That revealed the first growth loop:

better model → surprising demo → people share it → more people try ElevenLabs

Voice AI was especially suited to this.

You did not need to understand machine learning to see why the product was interesting.

You only had to hear it.

The free product turned every user into distribution

ElevenLabs opened its public beta on January 23, 2023.

People could generate speech, experiment with synthetic voices, and try voice cloning.

Giving users access early mattered because generated audio could leave ElevenLabs.

Someone could make a voice clip and post it on YouTube, X, TikTok, Reddit, Discord, or anywhere else.

The person listening did not have to visit ElevenLabs to experience the product.

The output was the demo.

This gave ElevenLabs a form of distribution that many SaaS products do not have.

When someone uses an analytics product, nobody else sees the dashboard.

When someone generates an unusual AI voice and puts it in a video watched by a million people, the product can spread with the output.

The loop looked like this:

someone discovers ElevenLabs → generates something surprising → shares it → other people hear it → some of them try ElevenLabs → they create more audio

ElevenLabs did not have to create every viral post itself.

Its users were creating demonstrations of the product every day.

Harry Potter wore Balenciaga

One of the clearest examples arrived only a few months after launch.

In March 2023, a creator published Harry Potter by Balenciaga.

The video imagined Harry Potter characters as models in a strange Balenciaga fashion campaign.

It combined Midjourney images, D-ID animation, and ElevenLabs-generated voices.

The video reached millions of views.

The New Yorker and Vox both covered it.

What mattered for ElevenLabs was not just the views.

The company had not planned the campaign.

ElevenLabs was simply one part of the creator's AI stack.

That pattern repeated.

Creators used AI voices for characters, faceless YouTube channels, memes, podcasts, audiobooks, games, and short videos.

Each new use case showed another group of people what the product could do.

By June 2023, only five months after opening its beta, ElevenLabs said it had passed 1 million registered users who had generated more than ten years of audio.

That was the first major stage of ElevenLabs' growth:

the product's output became its marketing.

Virality also created a safety problem

The same things that made ElevenLabs easy to spread also made it easy to misuse.

The voices sounded convincing.

Users did not need much audio to clone a voice.

The result could be shared anywhere.

Soon after the public beta launched, people started using the product to imitate celebrities and public figures.

The Verge documented people using ElevenLabs to generate hateful and misleading audio.

ElevenLabs responded by restricting some voice-cloning features, adding verification, improving traceability, and building detection and moderation systems.

The problem did not disappear.

As ElevenLabs grew, safety became part of the product itself.

This matters to the growth story because product-led distribution has a cost.

When the output leaves your platform, you lose some control over where it goes and how it is used.

ElevenLabs benefited from how easily its output spread.

It also had to deal with the consequences.

The API turned ElevenLabs from a tool into infrastructure

Creator virality brought ElevenLabs users.

The API gave it a different kind of growth.

Instead of making every customer use ElevenLabs directly, the company let developers put ElevenLabs voices inside their own products.

That changed the economics.

A creator might open ElevenLabs whenever they need to make a video.

A company that builds ElevenLabs into its app can generate usage every time one of its own customers does something.

The ElevenLabs API started appearing inside AI tutors, games, video tools, podcasts, accessibility products, companions, and other applications.

Now ElevenLabs could grow even when the person hearing its voice had never heard of ElevenLabs.

This created the second growth loop:

developer tries API → builds ElevenLabs into product → product gets users → users generate audio → API usage grows

ElevenLabs pushed this further with its Startup Grants program.

By August 2025, the company said it had given grants to more than 4,000 startups, unlocking more than 2.2 million hours of free AI voice generation.

The strategy helped both sides.

Startups got time to build without worrying as much about voice costs.

ElevenLabs got thousands of developers building products on top of its infrastructure.

If one of those startups grew, its ElevenLabs usage could grow with it.

Dubbing came back after the market changed

The original dubbing idea did not disappear.

It was just early.

By 2023, ElevenLabs had a better speech model, a growing user base, and much more evidence about how people wanted to use AI voice.

The company started expanding internationally.

In August 2023, ElevenLabs released Multilingual v2 with support for 29 languages.

Two months later, it launched AI dubbing that could translate spoken audio while preserving the original speaker's voice across more than 20 languages.

Now the dubbing product had something the original idea did not:

a proven model underneath it and customers already using ElevenLabs for audio.

This is another useful part of the company's growth strategy.

ElevenLabs did not have to choose between its original vision and the smaller product that found product-market fit.

It used the smaller product to earn the right to build the larger one later.

The Voice Marketplace created another growth loop

Voice cloning created a new problem.

If an AI company can reproduce someone's voice, what happens to the person whose voice it is?

ElevenLabs began turning part of that problem into a marketplace.

In February 2024, it introduced payouts for verified voice actors who shared professional voice clones in the Voice Library.

When paying customers used those voices, the creator could earn money.

This gave ElevenLabs another growth loop:

more voice creators → more voices → better selection → more users → more voice usage → more creator earnings → more voices

By May 2026, ElevenLabs said more than 10,400 creators were earning through the platform and total creator payouts had passed $22 million.

The marketplace made ElevenLabs more useful without ElevenLabs having to create every voice itself.

It also gave professional voice owners a reason to bring high-quality voices onto the platform.

That is more defensible than a text-to-speech model alone.

Models can get copied or catch up.

A marketplace gets stronger as both sides grow.

Voice agents unlocked a much larger market

The biggest shift came when ElevenLabs moved from generating speech to holding conversations.

Developers were already combining ElevenLabs with language models, phone systems, and company data to build voice agents.

But making a good voice agent requires more than text-to-speech.

The agent needs to know when someone has stopped talking.

It needs to respond quickly.

It needs to handle interruptions.

It needs to connect to company systems.

And it needs to sound natural enough that someone will stay on the call.

ElevenLabs began packaging more of that stack itself.

Voice agents opened use cases in customer support, sales, scheduling, recruiting, training, and other business workflows.

The market was suddenly much larger.

A company does not need to be a media company to need a voice agent.

Banks need phone support.

Telecom companies need customer service.

Retailers need sales and support.

Software companies need onboarding.

Governments need citizen services.

ElevenLabs said enterprise demand helped drive its jump from roughly $350 million ARR at the end of 2025 to more than $500 million by May 2026.

By September 2026, ElevenLabs said large organizations accounted for 55% of its revenue, up from 40% a year earlier.

That is a very different company from the creator-focused product that launched three years earlier.

But the two businesses are connected.

Self-serve let millions of people try the technology first.

Developers proved it could work inside real products.

Enterprise sales then helped large customers deploy it at much greater scale.

How ElevenLabs went from $0 to $600M ARR

ElevenLabs' revenue curve shows how quickly the business accelerated after finding product-market fit.

Staniszewski has said the company took roughly 20 months to reach $100 million ARR.

The next $100 million took around 10 months.

The company then reached roughly $300 million ARR about five months later.

ElevenLabs initially said it finished 2025 above $330 million ARR. A later company update described the year-end figure as roughly $350 million.

In February 2026, ElevenLabs raised a $500 million Series D at an $11 billion valuation.

Reuters independently reported the financing.

By May, ElevenLabs said ARR had passed $500 million.

In July 2026, CEO Mati Staniszewski said the business had reached roughly $600 million ARR.

These ARR numbers come from ElevenLabs and its CEO.

They should not be confused with audited recognized revenue.

But the direction is clear.

Revenue growth accelerated as ElevenLabs moved from individual creators into developers and then into large enterprise deployments.

The three growth loops behind ElevenLabs

Most explanations of ElevenLabs' growth focus on one thing:

the quality of its AI voices.

That was the starting point.

It was not the whole growth strategy.

ElevenLabs built three different growth loops over time.

1. The creator loop

A creator tries ElevenLabs.

They generate something interesting.

They publish it.

Other people hear the output and discover ElevenLabs.

Some of them become creators too.

This drove awareness and early user growth.

2. The developer loop

A developer tries the API.

They add ElevenLabs to their own product.

Their customers start generating audio.

As their product grows, ElevenLabs usage grows too.

This turned ElevenLabs into infrastructure.

3. The enterprise loop

Employees and developers become familiar with ElevenLabs.

A company finds a real business use case.

ElevenLabs helps it move from a prototype to a production deployment.

The company then expands voice AI across more teams, workflows, languages, or customer interactions.

This is now driving a large part of ElevenLabs' revenue.

The important part is the order.

ElevenLabs did not begin by cold-selling complex voice-agent contracts to giant companies.

It first made something individual people wanted to use.

What other startups can learn from ElevenLabs

Start narrower than the final vision

ElevenLabs wanted to transform dubbing.

The first product that found demand was much smaller: high-quality generated speech.

Solving that problem first gave the company the model, users, and distribution it needed to return to dubbing later.

Watch what users do, not only what they say

The early audiobook author was important because they were willing to use a terrible workflow to get the result they wanted.

That is stronger evidence than a survey response.

When users repeatedly force your product to do something it was not designed for, pay attention.

Build a product whose output can travel

ElevenLabs had an unusual advantage.

Its output could be copied, shared, embedded, and played anywhere.

Every impressive piece of audio had a chance to become a product demo.

The closer the thing users create is to an advertisement for the product itself, the easier organic distribution becomes.

Let developers distribute you

The API meant ElevenLabs did not have to own every customer relationship.

Thousands of other products could make ElevenLabs useful to their users.

That turned other companies into distribution.

Give suppliers a reason to join

The Voice Marketplace made voice owners part of the ecosystem.

As more voices joined, the product became more useful.

As more customers used those voices, creators earned more.

Good marketplaces give both sides a reason to make the other side stronger.

Move upmarket after people already want the product

Enterprise sales worked better because ElevenLabs was not introducing an unknown technology.

Employees, creators, and developers had already tried it.

Large companies could prototype before signing a large contract.

The enterprise team was turning existing demand into bigger deployments rather than trying to create demand from nothing.

The real reason ElevenLabs grew

ElevenLabs did not grow because of one viral video, one launch, or one marketing channel.

It kept turning product usage into new forms of distribution.

At first, the voices themselves spread.

Then developers put those voices inside other products.

Then voice creators expanded the supply.

Then enterprises took the same technology and used it for millions of customer interactions.

The company kept widening the market without abandoning the thing that made the first product work:

audio that sounded good enough for people to want to use it.

That is the part of the ElevenLabs story worth copying.

Build something people want to show other people.

Watch what they force it to do.

Then build around the behavior that keeps repeating.

Frequently asked questions

How did ElevenLabs grow?

ElevenLabs grew through a mix of product quality, creator sharing, a free self-serve product, API adoption, multilingual speech, a voice marketplace, and enterprise voice agents.

Its early growth came from creators sharing AI-generated audio. Developers then embedded ElevenLabs into other products, while enterprise voice agents became a major revenue driver.

How did ElevenLabs get its first users?

ElevenLabs first built interest through voice demos, newsletters, technical communities, and a waitlist.

CEO Mati Staniszewski said one early demo brought roughly 1,000 people onto the waitlist the next day.

The public beta then made the product easy for creators and developers to try and share.

What was ElevenLabs' main growth strategy?

ElevenLabs did not rely on one growth strategy.

Its first loop was creator-led: users generated audio and shared it publicly.

Its API created a developer-led loop, where other products generated recurring usage.

Later, enterprise sales expanded the technology into large support, sales, and operational deployments.

When did ElevenLabs launch?

ElevenLabs launched its public beta on January 23, 2023.

It initially focused on text-to-speech and narration while continuing to work toward its original goal of AI dubbing.

How many users did ElevenLabs have in its first year?

ElevenLabs said it passed 1 million registered users by June 2023, around five months after launching its public beta.

What is ElevenLabs' latest ARR?

CEO Mati Staniszewski said ElevenLabs had reached roughly $600 million in ARR by July 2026.

The number is company-reported and should be treated as annualized recurring revenue rather than audited recognized revenue.

What is ElevenLabs worth?

ElevenLabs raised $500 million at an $11 billion valuation in February 2026.

Reuters independently reported the Series D financing.