ElevenLabs launched Eleven v4 and v4 Turbo on September 28, adding support for 90-plus languages and targeting lower-latency voice-agent use cases. TechCrunch reported the company also claims voice cloning can work from roughly 10 seconds of source audio.
The bigger product shift is control. ElevenLabs says v4 can follow more detailed inline direction, including stacked expression tags, while preserving context across longer passages. That could cut down on splitting a response into multiple scripted segments just to manage tone, pacing, or emphasis.
There is also an early benchmark win in English, but not a blanket lead across every language. The edge cited in coverage is specific to an English voice arena, while separate reporting notes stronger competition on some non-English boards. For teams evaluating a switch, the practical checks are still latency under load, pronunciation quality in target languages, and whether cloned voices remain stable in production. Current plan details are on the ElevenLabs API pricing page.