Eleven V4

What Eleven V4 is good at, and where it isn't

Eleven V4 is ElevenLabs' answer to the hardest part of synthetic voiceover: performance. Most text-to-speech reads the words correctly; few read them like an actor. ElevenLabs released Eleven v4 in September 2026 on a new architecture and calls it its most emotive model yet.

The flagship capability is direction inside the script. Write tags like [laughs], [whispers] or [sighs] exactly where they should happen, and v4 performs them in order, more reliably than v3, so a single line can start calm, break into a laugh and drop to a whisper. The emotional range carries across more than 90 languages, which makes it practical to localize the same performance for several markets with the same voice.

It comes in two versions. Standard is the most expressive and the right choice for finished voiceovers. Turbo keeps nearly the same delivery, returns audio faster and costs much less, which makes it the sensible default for long narration, drafts and high-volume work. On ZOOOP you choose from 20 curated voices and set stability: lower for a livelier, more varied read, higher to keep every line of a long script consistent.

Where it's weaker: each generation takes up to 5,000 characters, so long scripts have to be split into sections. There is no separate style field; direction goes into the text as tags. If you would rather keep the script clean and describe the delivery separately ("warm and slow", "a dramatic newscast"), Gemini 3.8 Flash TTS is built that way. Voice cloning is not available on Eleven V4 on ZOOOP.

A reasonable mental model: make Eleven V4 the default for narration and character voices, Standard for the final mix and Turbo for drafts and long reads. Reach for Gemini 3.8 Flash TTS when you want style instructions, Multilingual V3 to match an existing v3 project, and Kling Avatar V2 to turn the voice into a talking video.