ElevenLabs Eleven v4 is an expressive text-to-speech model for creators and enterprises who need natural, emotive voice audio. Launched on September 28, 2026, it succeeds Eleven v3, which reached general availability in March 2026.
Core Features
- A new architecture that reads tone, pacing, emotion, and context to deliver dramatic, tender, urgent, or conversational speech.
- Ranks number one of 92 models on the Artificial Analysis leaderboard, 43 Elo points clear of second place.
- Clones a voice from as little as 10 seconds of audio, with Instant and Professional voice-clone options.
- Supports more than 90 languages with native accents, including improved Japanese, Brazilian Portuguese, Mandarin, and Cantonese.
- Inline audio tags, multi-speaker dialogue, and a v4 Turbo variant at about 100 milliseconds median latency for live agents.
Use Cases
- Audiobook producers and YouTube creators who need consistent, characterful narration.
- Game and animation studios building voiceovers from a few reference clips.
- Conversational AI agents and live customer service using v4 Turbo's streaming latency.
Pricing
A free tier includes 10,000 credits per month. Paid plans run Starter at 6 dollars, Creator at 22 dollars (first month 11 dollars), Pro at 99 dollars, Scale at 299 dollars, and Business at 990 dollars. API pricing is 0.08 dollars per 1,000 characters for v4 and 0.04 dollars for v4 Turbo. ElevenLabs raised a 500 million dollar Series D at an 11 billion dollar valuation in February 2026.
Pros and Cons
Pros: best-in-class voice quality and cloning, with broad language coverage.
Cons: the credit system frustrates some users, and a BIPA class-action lawsuit over voice training is pending.
Our Take
Best for projects where voice quality directly drives revenue. The trade-off is cost and the credit model, which can exceed advertised rates on retries. See more in our AI audio tools category.
