Launching today
Eleven v4 and Eleven v4 Turbo
Elevenlabs' fastest and most emotive voice models yet
74 followers
Elevenlabs' fastest and most emotive voice models yet
74 followers
Meet Eleven v4 and Eleven v4 Turbo by ElevenLabs, their most expressive models yet, with Turbo built for real-time use. Available in apps, API, and agents.
Eleven v4 and Eleven v4 Turbo are ElevenLabs’ fastest and most emotive voice models yet.
Built on an entirely new architecture, Eleven v4 interprets tone, pacing, emotion, character, and context to generate speech that feels performed. Eleven v4 Turbo brings the same research to low-latency use cases such as agents, generating speech with a median inference latency of ~100 ms.
Voice cloning is more authentic and consistent. Instant Voice Clones capture voices with extraordinary fidelity from just 10 seconds of audio, and Professional Voice Clones offer our highest fidelity. For long-form creation, Eleven v4 preserves speaker identity reliably across infinite generations, perfect for long form narration, dialogue, and regenerated lines.
Direct the performance with inline tags. Specify delivery, emotion, pacing, reactions, SFX, or style. Eleven v4 stages the scene and performs it with naturalness and precision.
What's new:
A completely new Text to Speech architecture built for expressive performance
Eleven v4 Turbo for agents and anything live, at a 100ms median latency
Instant Voice Clones from 10 seconds of audio, Professional Voice Clones for highest fidelity
Improved performance across 90+ languages, with new languages including Cantonese, Mongolian, and Odia
Speaker identity preserved across long-form narration, dialogue, and regenerated lines
More natural multi-speaker dialogue
Pronunciation control with IPA support
Inline tags for direct control over delivery, emotion, pacing, reactions, SFX, and style
Available now in ElevenCreative, ElevenAgents, and ElevenAPI.
Try Eleven v4 → https://elevenlabs.io/v4
P.S. I hunt the latest and greatest launches in tech, SaaS and AI, follow to be notified → @rohanrecommends
@rohanrecommends this update looks impressive especially the focus on more natural delivery voice cloning and low latency for real time use cases like ai agents
@rohanrecommends Great job, guys! This is really impressive. This time it's almost impossible to recognize whether it's an AI voice or a human.
v4 with tags works really well ngl
The ~100 ms median latency on Turbo is the number that matters for us. We build a voice companion that phones older adults every day, and for that audience the pause after they finish speaking is what breaks the feel of a conversation, since many speak slowly and with long gaps. Curious how v4 handles hesitations and trailing-off sentences when used in agents: do the inline tags let you slow pacing and soften delivery per turn, or is that inferred from context?