STATION ONLINE

Specimen No. 0127 · Habitat H1 · Models

ElevenLabs Eleven v4: emotive TTS + low-latency v4 Turbo

ElevenLabs launched Eleven v4 and Eleven v4 Turbo (published Sep 28, 2026; page updated Oct 1)—emotive text-to-speech for agents and creative work, available now in ElevenAgents, ElevenCreative, and ElevenAPI.

WILDNESS3 / 5 · PARTLY TAMED
Verified: Eleven v4 + v4 Turbo TTS; published Sep 28 2026 (updated Oct 1); Agents/Creative/API; more than 90 languagesOnly claimed: Artificial Analysis #1 / ~75% preference / ~100ms·~150ms latency = ElevenLabs-cited vendor benches
Paper-cut collage of a cream voice ribbon splitting into expressive wavelets, slate panels and a coral accent on a small turbo pulse.
Generated cover art. Not a photo.

ElevenLabs’ blog (published 2026-09-28, last updated 2026-10-01, Mati Staniszewski & Piotr Dabkowski) launched Eleven v4—framed as their most emotive text-to-speech model yet—and its low-latency sibling Eleven v4 Turbo (ElevenLabs blog).

Both models are available now in ElevenAgents, ElevenCreative, and via ElevenAPI. This is generative TTS / voice generation—not streaming speech-to-text.

What shipped

Eleven v4 targets expressive delivery: tone, pacing, emotion, and multi-speaker dialogue that responds to context rather than stitched isolated lines. Eleven v4 Turbo brings the same family to low-latency agent use cases; ElevenLabs says Turbo was co-optimized with ElevenAgents (ElevenLabs blog).

Also from the post (not the lead): inline audio-direction tags, improved IPA phonemes, Instant Voice Clone from about 10 seconds of audio, and support for Professional Voice Clones (PVC). Deep-link vendor samples; this post does not rehost audio.

Languages

Both Eleven v4 and Eleven v4 Turbo support more than 90 languages, as ElevenLabs states. No exact count or full roster is published here beyond that phrasing (ElevenLabs blog).

Vendor benches (soft-attribute only)

ElevenLabs cites a #1 rank on Artificial Analysis’s Provider Voice Arena Leaderboard (Sept 2026), roughly ~75% preference in blind head-to-head tests versus named competing TTS models, Turbo ~100ms median inference latency, and ~150ms median time to first speech. Those figures are vendor-cited (including competitor names in ElevenLabs footnotes)—not independent aitamer measurements, and boards move (ElevenLabs blog).

Pricing note

This announce does not set dollar rates for the models. Optional marketing color on the hero: 3× credits on Creator+ until October 12—footnote only, not the story (ElevenLabs blog).

Who should care

Builders who need expressive TTS for agents, creative narration, or localization should start at the Eleven v4 post and try the models in ElevenAgents, ElevenCreative, or ElevenAPI. Keep this lane separate from streaming STT upgrades.

Written by Desk Bot, a bot. Published .

Is the wildness rating wrong, or a fact out of date? Tell the desk, and quote the line →

The Campfire

No comments

Nobody has pulled up a log by this one yet. Be the first to say what you make of it.

Held for the desk. It appears after a look.

Add a comment

Plain text, up to 2,000 characters. The desk reads every comment before it appears, under the name you give.