Eleven Speech is ElevenLabs’ text-to-speech model on Mage. It reads your text in one of 16 voices, in more than 90 languages, and takes direction from audio tags written in square brackets. Its current model is Eleven v4.
Overview
Unlimited generation by plan
Eleven Speech is priced by the length of the text it speaks, audio tags included: 30 Gems for each started block of 250 characters.
What it supports on Mage
Best for
- Narration, voiceover, and character lines that need real acting
- Speech in languages other than English
- Sound cues placed inside the speech
Write a tag in square brackets before the line it should change. Tags can set emotion ([excited], [sarcastic]), delivery ([whispers], [shouting]), reactions ([laughs], [sighs]), and sound cues ([applause], [door slams]). Be specific: [low, gravelly voice] works better than a vague tag, which the model may read as a sound effect.
Stability controls how much the performance varies: lower values are more expressive, higher values keep the delivery steady.
Limitations
- It speaks in its preset voices only; it does not take a voice from an audio reference yet. For a voice that should recur across video clips, use a Character Voice.
- Speech only. For songs, use Eleven Music.
Use via API
Eleven Speech is available through the Mage API as the eleven_speech architecture. Its fields, tokens, defaults, and price are on its API reference page. Last modified on October 24, 2026
