ElevenLabs releases Eleven v4 and Eleven v4 Turbo text-to-speech models

On September 28, 2026, ElevenLabs released Eleven v4, which it calls its most emotive text-to-speech model, and the lower-latency Eleven v4 Turbo for agents. Both support more than 90 languages.

Archive entry: this covers an event from September 28, 2026, and we published it on November 8, 2026. The details come from sources dated at the time and may have changed since.

On September 28, 2026, ElevenLabs released Eleven v4, which it calls its most emotive text-to-speech model, together with Eleven v4 Turbo, a lower-latency version for agents. The company described both in its announcement on the ElevenLabs blog.

What ElevenLabs released

Eleven v4 is built on an entirely new architecture, and Eleven v4 Turbo is the lower-latency version for agent use.
Both models support more than 90 languages, and a voice recorded in one language can speak others with a native-speaker accent.
Eleven v4 added support for Professional Voice Clones (PVC), and Instant Voice Clones reach high fidelity from just 10 seconds of audio.
It follows audio tags and direction prompts, including inline tags such as [laughs] or [phone buzzing], more accurately than prior models.
Improved IPA phoneme support makes custom pronunciations more reliable, and request stitching is significantly more reliable.
Eleven v4 Turbo has a median inference latency of about 100ms and a median time to first speech of about 150ms, in ElevenLabs tests from September 2026 that exclude network latency.

What it means for people who build with AI

For people building narration, dubbing or voice agents on ElevenLabs, the release added finer control through audio tags and direction prompts, Professional Voice Clone support and a Turbo variant aimed at low-latency agent use. Because both models supported more than 90 languages and a voice could speak languages other than the one it was recorded in, the same voice could be reused across languages.

Availability and cited results

Both models were available in ElevenAgents, ElevenCreative and through ElevenAPI at launch. The post did not state pricing or API model IDs. ElevenLabs cited a first-place ranking from Artificial Analysis (Provider Voice Arena Leaderboard, September 2026) and said that about 75% of listeners preferred v4 in blind head-to-head tests.

Sources

Introducing Eleven v4, our most emotive model (ElevenLabs, September 28, 2026, official)

Written with AI assistance, strictly from the sources listed above. Before it went live, automated checks verified its dates, sources and every language edition.