ElevenLabs on Mage

Two. Eleven Speech is ElevenLabs' text to speech, running Eleven v4, its most expressive voice model: 16 voices, more than 90 languages, and audio tags such as [whispers], [laughs] or [door creaks] that direct the read line by line. Eleven Music is ElevenLabs' music model: full songs with lyrics and vocals, or instrumentals, from 10 seconds to 5 minutes.

Both run through ElevenLabs itself, so the voices and the songs are the ones ElevenLabs makes. You use them with your Mage account instead of a separate ElevenLabs subscription: Gems on any plan, or unlimited on a paid plan up to a length each plan covers.

Mage is an independent creative platform, not part of ElevenLabs. What it adds is everything around the audio: image models such as Mango 3 and Nano Banana, video models such as Cherry 2 Pro, saved characters that keep their face and voice across clips, and apps like Lipsync and Music Video that put the audio on screen.

Eleven v4 reads the whole script before it speaks, so its stress, pacing and emotion follow the meaning. Tags in square brackets direct it: [excited], [sarcastic] and [softly] change the emotion; [whispers] and [shouting] the volume; [sighs] and [laughs] add a reaction; [door creaks] or [applause] add a sound in the speech. The MP3 is exactly as long as the reading, and one request takes up to 5,000 characters.

Pick from 16 voices, from a warm storyteller to a polished British anchor, and write in any of more than 90 languages. The AI text to speech guide covers scripts and voices in depth, and the Eleven Speech model page has samples in English, Spanish, French and Japanese.

• Put a tag right before the line it changes
• Use ellipses for pauses
• Lower Stability for a more expressive read
• Write numbers as words when the reading matters

Eleven Music composes a new song from a description, writes its lyrics and sings them, or plays an instrumental when you turn Instrumental on. Lengths are exactly 10, 30, 60, 120, 180 or 300 seconds. Describe the genre, mood, tempo, instruments, singer and subject; name a style rather than an artist, because prompts that name a real artist or quote existing lyrics are refused.

For more on writing song prompts, read the AI song generator guide. The Eleven Music model page has samples from pop and R&B to reggaeton and lo-fi.

On a paid plan, a request runs without Gems when it fits the plan. Eleven Speech is measured by the length of the text: Basic covers 750 characters per request (about 50 seconds of speech), Pro 1,500, Pro Plus 3,250, and Max 4,750. Eleven Music is measured by the length of the song: Basic covers 10-second songs, Pro 30 seconds, Pro Plus one minute and Max two minutes. Anything longer runs on Gems, at 30 Gems per started block of 250 characters for speech and 3.75 Gems a second for music.

Compare the plans on the membership page.

Developers call POST /v1/eleven_speech/generate and POST /v1/eleven_music/generate on the Mage API, with one key and one Gem balance for every Mage model. The Mage API page covers keys and prices. Connect the Mage MCP server, and Claude, ChatGPT and other assistants can speak a script or write a song for you in the chat, quoting the Gems first.