Eleven V4 Turbo by ElevenLabs

Overview

Eleven V4 Turbo is an audio model from ElevenLabs for expressive text-to-speech. It reads your script like a performance, then generates spoken audio that matches tone and pacing cues you write inline. You can add bracketed audio tags to steer delivery at the phrase level. This is useful when you need speech that sounds acted, not read.

What you can build

• Real-time voice agents for support, sales, or appointment booking
• Interactive game characters with fast back-and-forth dialogue
• Expressive voiceovers for short-form video and ads
• Multi-turn assistant voices that keep a consistent character
• Localized narration that keeps the same voice identity across languages

Inputs

• A script to speak, as plain text, up to 10,000 characters.
• Optional inline delivery cues inside square brackets, like [whispering], [laughing], [shouting], or [excited].
• A built-in voice selection from: Adam, Alice, Bella, Bill, Brian, Callum, Charlie, Chris, Daniel, Eric, George, Harry, Jessica, Laura, Liam, Lily, Matilda, River, Roger, Sarah, Will.
• An MP3 output choice that sets sample rate and bitrate. Supported options are 22.05 kHz at 32 kbps, or 44.1 kHz at 32, 64, 96, or 128 kbps.
• An optional stability control from 0 to 1. Lower values allow more variation and emotion. Higher values keep delivery steadier.
• An optional similarity control from 0 to 1. Higher values track the target voice more closely. This can reduce naturalness.

Outputs

The model returns a single MP3 audio file of the spoken script. The file uses the sample rate and bitrate you selected. The content is one continuous speech track with timing shaped by your punctuation and inline tags.

Limitations

• One request is limited to 10,000 characters.
• SSML break tags are not supported for Eleven v4 family models. Use punctuation and audio tags to shape pauses.
• Tag control is strong, but it is not perfect. Some tags may be ignored or softened.
• Messy input can hurt pacing. All-caps text, missing punctuation, or inconsistent tag style can sound unnatural.
• This Wiro model exports MP3 only. If you need PCM, Opus, or telephony codecs, use a different workflow.

Safety & compliance

Follow ElevenLabs’ Prohibited Use Policy and safety rules when generating speech.

• Only generate voices you have rights to use. Get clear consent for any cloned voice.
• Don’t use the model for deceptive impersonation, fraud, or harassment.
• Expect safeguards around high-risk voices. Some cloning attempts may be blocked.
• Generated audio may include traceability features. Plan for disclosure where your platform or law requires it.