Skip to content
by InfinyAI ES

IA en un minutoNewsModels

ModelsUpdated:

ElevenLabs launches Eleven v4, its most expressive voice model

IA en un minuto newsroom · Editor: Jon Elgezabal

In 30 seconds

Eleven v4 is the new ElevenLabs text-to-speech model, and it lets users mark emotion, delivery or even sound effects with tags placed inside the text. Its Turbo version, built for voice agents that cannot keep callers waiting, starts answering in about 150 milliseconds. Both models speak more than 90 languages, Spanish and Catalan among them, and instant cloning needs only 10 seconds of recording.

On September 28, ElevenLabs released two text-to-speech models: Eleven v4, which it describes as the most emotive it has built, and Eleven v4 Turbo, a low-latency variant aimed at voice agents. According to ElevenLabs, anyone building such an agent has so far had to choose between speed and expressiveness, because higher-quality models were slower to produce speech.

Eleven v4 runs on a new architecture and lets the delivery be directed from the text itself. Inline tags in square brackets, for a laugh, an accent or a buzzing phone, for instance, set how certain phrases should be said, what emotion to convey and even which sound effects to add. The company says conversations between several speakers and custom pronunciation through the International Phonetic Alphabet have improved as well.

Eleven v4 and its Turbo version cover more than 90 languages, Catalan and Spanish among them. According to the ElevenLabs blog, a voice recorded in one language can now speak any of the others with a native accent while keeping its identity. Instant Voice Clones needs only 10 seconds of audio to capture a voice, and Eleven v4 also adds Professional Voice Clones.

The Turbo version was tuned together with ElevenAgents, the company's conversational agent platform. It takes a median of about 150 milliseconds to start speaking, a figure ElevenLabs measured in September over WebSocket streaming with network latency taken out. In blind tests that same month against models from Cartesia, Inworld and Google, around 75% of listeners preferred Eleven v4, which the company says also holds first place on the Artificial Analysis voice leaderboard.

Both models can already be used inside ElevenAgents and ElevenCreative and through ElevenAPI.

Why it matters · analysis and opinion

Choosing between a fast voice and a nuanced one was, according to ElevenLabs, the price of any phone agent, and Eleven v4 Turbo aims to remove it. For anyone who answers calls or produces audio, the practical change lies in the tags: tone, emotion and even a sound effect are written into the script, the way a director would guide a voice actor. Having Spanish and Catalan among the supported languages opens the door to businesses that work in those languages without going through English. One serious caveat remains: the blind tests that favor Eleven v4 were organized by the company itself. Before putting an agent like this on the phone, it is worth listening to it with real street names, brands and surnames, and telling callers that a machine is on the other end.

Official source: ElevenLabs · Written with the help of AI: how we make the news

AI in your inbox, every day or every Friday

The stories that matter, each one in a minute. With the source for every one.

Choose one or both:

Sign up and you are in: the daily arrives every night and the weekly on Friday mornings. You can unsubscribe from any email.