ElevenLabs was founded to fix dubbing. It shipped narration first.

On 5 November ElevenLabs said people can now publish songs made in ElevenMusic straight to SoundCloud, TechCrunch reported. No download, no re-upload. The songs go up labelled as AI-made, according to The Verge. ElevenLabs also gave a number: users have made over 25 million tracks with its models.

That number counts songs, and songs were never the plan. In September 2022 the company published a blog post on voice conversion that named its "chief product". That product wasn't music, and it wasn't narration either. It was "the identity-preserving automatic dubbing tool", due "early next year."

The dubbing tool shipped on 10 October 2023. Narration came first. Here is why.

The problem they set out to fix

Mati Staniszewski and Piotr Dąbkowski grew up in Poland. There, a foreign film is often voiced in Polish by one narrator for every part. On Sequoia's Training Data podcast in July 2025, Staniszewski described it: "whether it's male voice or female voice—still narrated with one single character."

The idea came in late 2021. Dąbkowski was about to watch a film with his girlfriend. She didn't speak English, so they put it on in Polish.

So the plan was dubbing. Take a video, keep each speaker's own voice, change the language.

Creators said: one day

In 2022 they raised a $2 million pre-seed. ElevenLabs says that first round was priced at a $9 million valuation. Then they tested the idea on creators.

Nathan Benaich interviewed Staniszewski on stage at RAAIS, an AI summit in London, and wrote up the talk. The team emailed over a thousand YouTubers with demo samples. About five were interested. And those few did not want full dubbing. They wanted better voiceovers and cleaner narration.

Staniszewski told David Senra in September what creators kept saying: "Great, I would love to get dubbing one day, but today I have different problems." The problems were small. Let me fix a line I recorded wrong. Let me hear my script before I record it. Have AI speak over the video instead of me.

He gives the research as the first reason. "All of the components to do dub were not good enough yet," he said. So they split the work. "Before we can solve dubbing, let's solve the research component to generate speech and make it sound great." On the product side, let people "just narrate content," and set the language shift aside for now.

The narration creators asked for was the first part of the dubbing tool, so serving them now was also building toward the plan. When users say "one day," ask what they need today. Then check whether it's a piece of the thing you meant to build anyway.

A tweet-sized text box and a whole book

The beta ran in late 2022 with a small group of users. Many were book authors. The product was "literally a small text box," Staniszewski told Sequoia. "It was like a tweet length, effectively."

One author pasted his entire book into it. Benaich's account says he generated it section by section and stitched the audio into an audiobook. Most platforms banned AI content at the time. He uploaded it anyway. "They thought it was human," Staniszewski said. The book got good reviews. The author came back with friends who were also authors, saying they needed it.

Benaich quotes Staniszewski's reaction: "Oh damn, we might be onto something."

ElevenLabs launched publicly in early 2023. Next came creators running "no face" channels, where the creator never appears on screen and a voice narrates. Staniszewski said the trend spread "like wildfire" in the first six months, and ElevenLabs supplied the voices for a lot of it.

None of that was dubbing. All of it ran on the speech engine dubbing would need.

The original plan, most of a year late

AI Dubbing arrived on 10 October 2023. The launch post said it was "born from the foundational mission" ElevenLabs was built on. It translated speech into more than 20 languages, Polish among them, and kept the original speaker's voice.

Then users stretched that product too. Someone tried to dub singing videos, which the model was never built for. "It kind of didn't work," Staniszewski told Sequoia, "but it gave you, like, a drunken singing result." He said the clips went viral a few times.

In December 2025 dubbing got its biggest-name partner: Meta. Staniszewski posted that ElevenLabs and Meta were "starting with the very use case that inspired ElevenLabs in the early days: dubbing." The deal covers translating Instagram Reels into many languages.

The cycle he still runs

On the Sequoia podcast Staniszewski described the method that came out of those years. Release each new model to "the entirety of the user base". Learn from what people do with it. "The prosumer adoption is there, and then the enterprise adoption follows with additional product, additional reliability." Then a new model ships, "and kind of the cycle repeats."

The revenue shows the enterprise half growing. In September 2025 ElevenLabs said it had passed $200 million in annual recurring revenue (ARR) that year. It said it was "rapidly approaching a 50/50 revenue split" between enterprise and self-serve customers. On 30 September 2026 it said enterprise was 55% of revenue, driven, the company says, by its voice-agents product. That came with a $300 million employee tender at a $22 billion valuation. The first round, four years earlier, priced the company at $9 million.

Music started with the labels

Music ran the same cycle, with one change: the labels came in before launch, not after. Eleven Music launched on 5 August 2025. ElevenLabs said it was made "in collaboration with labels, publishers, and artists" and cleared for nearly all commercial uses. According to TechCrunch, a consumer app followed in April 2026 and a new music model a month later.

The enterprise-style step came after. In September 2026, Universal Music Group said it was building an AI remix platform on its catalogue through a multiyear licensing agreement with ElevenLabs. Artists choose whether to take part.

We found no interview where Staniszewski links the drunken-singing dubs to Eleven Music, and we're not claiming they did. What the record does show is the order of things. The 2022 plan put dubbing first. The research had to start with speech, and creators asked for narration, so narration shipped. Dubbing came on top of it, and music came later, with labels in the room from the start.

The September 2022 post naming dubbing as the chief product is still on ElevenLabs' blog. This week the company's headline number counted songs.