ElevenLabs Models: The Essentials

Last updated: September 29, 2026

ElevenLabs builds most of Scenario's audio models: full songs, expressive speech, voice transformation, dubbing, and sound effects. This article covers all fifteen ElevenLabs models on the platform, led by the new v4 and v4 Turbo speech models, with an example for each: Music v2.5, Music Advanced v2.5, Music Extend v2.5, Music v2, Music Advanced v2, 3 (Alpha), Multilingual v2, Turbo 2.5, Voice Changer, Voice Isolator, Speech to Speech, Dubbing v2, Dubbing, and Sound Effects 2.

New: ElevenLabs v4 and v4 Turbo are now on Scenario. They are the most expressive ElevenLabs voices yet: write the script, add audio tags like [whispering], [laughing] or [shouting], and the voice performs it in 100+ languages. Jump to Text to Speech below.

The Family at a Glance

Sections below follow this order, and inside each section the newest model comes first.

Category	Models (newest first)	Use it for
Text to Speech	v4, v4 Turbo, 3 (Alpha), Multilingual v2, Turbo 2.5	Voiceovers, narration, dialogue, voice agents
Music	Music v2.5, Music Advanced v2.5, Music Extend v2.5, plus legacy v2 models	Full songs from a text prompt
Dubbing	Dubbing v2, Dubbing	Translating speech while keeping the voice
Voice Transformation	Voice Changer, Voice Isolator, Speech to Speech	Changing or cleaning up a voice recording
Sound Effects	Sound Effects 2	Custom sound effects from text

Text to Speech

Convert written text into spoken audio, choosing a built-in voice or a cloned one. The newest models, v4 and v4 Turbo, come first.

ElevenLabs v4 (New)

The most expressive ElevenLabs voice model, and the new default for voiceovers, narration, audiobooks and character dialogue. Steer the delivery with inline audio tags, either single words like [sighs] or free-form directions like [gruff old sailor, thick Scottish accent]. 100+ languages, 21 built-in voices or your own clone, up to 10,000 characters per take. Lower stability (0.15 to 0.35) for emotion and shouting, raise it (0.7 to 0.8) for narration and numbers.

"[exhales] Okay. Okay, so... I have to tell you something..." Comic timing from tags alone · Open on Scenario

"The Last Ferry": a narrator, a Scottish sailor, a young woman and a whispering ghost, all from one voice · Open on Scenario

A breakup call that moves from whisper to tears to fury to resignation (stability 0.15) · Open on Scenario

One voice greeting in Hindi, Arabic, Korean, Mandarin, Swahili and Turkish · Open on Scenario

ElevenLabs v4 Turbo (New)

The same voices, tags and languages as v4, generated 2 to 4 times faster. Use it to draft scripts, prototype game dialogue and power voice agents, then re-run the final take on v4 with the same text, voice and settings.

"[excited] Okay, quick update, team!..." A standup update with a laugh, a whisper and a sigh · Open on Scenario

A controller and a pilot trade lines, with an alarm and dry humor, from a single voice · Open on Scenario

A Korean weather bulletin with tags written in Korean · Open on Scenario

Full guide with 27 more examples and every prompt: ElevenLabs v4: The Essentials.

Earlier text-to-speech models:

ElevenLabs 3 (Alpha)

The previous expressive generation, now succeeded by v4: 70+ languages, steered with inline tags like [whispers] or [excited] for emotion, pacing, and emphasis.

"[low][dark] Welcome... to my domain." Inline emotion tags steering delivery · Open on Scenario

ElevenLabs Multilingual v2

Natural, emotionally rich speech in 29 languages, with 21 built-in voices or a cloned one.

Greek: "Είμαι καλά! Θέλεις να πάμε μια βόλτα το απόγευμα;" · Open on Scenario

ElevenLabs Turbo 2.5

Low-latency text to speech in 32 languages, tuned for near real-time use.

“Don't make a sound... please.” Whispered delivery at low latency · Open on Scenario

Music

Full tracks generated from text. Music v2.5 takes one prompt and writes a whole song. Music Advanced v2.5 builds a song section by section, with its own styles, lyrics, and duration per part. Both are the current recommended versions; Music v2 and Music Advanced v2 remain available but are superseded. For a full walkthrough of prompting either v2.5 model, see the dedicated "ElevenLabs Music v2.5: The Essentials" article.

ElevenLabs Music v2.5

One text prompt in, one complete track out: any genre, vocal or instrumental, 3 to 600 seconds.

Uplifting pop anthem, vocal chorus · Open on Scenario

Warm neo-soul groove with vocal ad-libs · Open on Scenario

Synthwave chase, instrumental · Open on Scenario

5-second brand sting, the shortest supported length · Open on Scenario

5-minute long-form cinematic score, near the 600s maximum · Open on Scenario

ElevenLabs Music Advanced v2.5

Compose a song as an ordered list of sections, each with its own styles, lyrics, duration, and a context-adherence dial.

Pop anthem, verse and chorus · Open on Scenario

ElevenLabs Music Extend v2.5

Reworks an existing track instead of starting from a blank prompt: keep the sections you want unchanged and generate new intros, verses, bridges, or outros around them. For the full parameter reference and more examples, see the dedicated "ElevenLabs Music v2.5: The Essentials" article.