ElevenLabs exits beta with 28-language voice model

Imagine recording a brief audio clip in your native tongue, and having an AI instantly turn around and speak fluent Japanese, Classic Arabic, or Finnish using your exact vocal timbre, tone, and pacing.

Voice AI startup ElevenLabs has officially graduated from its beta phase, marking the milestone with the release of Eleven Multilingual v2, a new foundational speech model supporting 28 languages.

According to the company, which launched its platform in January, its global user base has already crossed the one million mark.

One voice across nearly 30 languages

The headline feature of Eleven Multilingual v2 is its ability to automatically detect input text across nearly 30 written languages and generate nuanced, emotionally expressive speech on the fly.

The update adds a massive batch of new languages, including Chinese, Japanese, Korean, Dutch, Turkish, Swedish, Indonesian, Filipino, Ukrainian, Greek, Czech, Finnish, Romanian, Danish, Bulgarian, Malay, Slovak, Croatian, Classic Arabic, and Tamil.

These join the existing roster of English, Polish, German, Spanish, French, Italian, Hindi, and Portuguese.

The real breakthrough lies in voice preservation. Regardless of whether a creator uses a pre-made synthetic voice or a cloned voice, the unique vocal characteristics, including the speaker's original accent, carry over intact across all 28 languages.

Paired with the company's recently rolled out Professional Voice Cloning tool, users can effectively produce a digital replica of their own voice that speaks dozens of languages they have never studied.

ElevenLabs says the underlying technology was developed entirely in-house over 18 months, focusing on context understanding and emotional delivery. The company also announced plans to launch a voice-sharing marketplace where creators can share voices and earn revenue.

Why it matters

For solo creators, indie game developers, and small publishers, localization just got radically cheaper. Translating a narrative video game or an audiobook into two dozen languages traditionally requires hiring dozens of voice actors and audio engineers, a luxury only massive studios could afford.

Now, a developer can localize an entire script into 28 languages while keeping the same actor's voice acting intact across international releases. In our view, this levels the global distribution field in a very practical way.

For the voice AI sector, ElevenLabs is shifting the benchmark. Competitors can no longer get away with building great English-only models with a few robotic translations tacked on. The standard is now foundational multilingual delivery that carries emotional nuance and persistent identity.

That said, the technology comes with clear risks. As realistic voice cloning becomes multilingual and effortless, malicious impersonation and automated disinformation cross language borders just as easily. While ElevenLabs has introduced safety measures alongside this launch, the pressure on its safeguards will multiply as it rolls out voice-sharing monetization to over a million users.