Noesion Radarnoesion.ai
Release2026-09-28

ElevenLabs launched v4 and v4 Turbo speech models. They add more expression control and support for 90 languages. Voice cloning now needs only 10 seconds of audio. The model handles voice identity better over long text. It uses inline tags to define and stack expressions. Latency is lower for voice agents. Audio generation starts as soon as the LLM begins answering. The model improves handling of confrontations and escalations. Japanese, Brazilian Portuguese, Mandarin and Cantonese saw the biggest quality gains.

What changed

v3 supported 70 languages and required longer audio clips for voice cloning.

What it unlocks

Cloning a voice with a 10-second audio clip and stacking multiple expression tags.

  • 90 languages supported
  • 10 seconds for voice cloning
  • 70 to 90 languages

Sources

Send this to someone who needs it

Shares the story and its sources. Nothing about you.

What does this mean for your job?

This is the story as everyone gets it. Once a week we send you the version written for your role — what changed, why it matters for the work you actually do, and one thing to try. Free while we tune it.