Voice · guidance updated 4 Oct 2026

ElevenLabs prompt guide

Eleven v4 (eleven_v4; eleven_v4_turbo for real time) is the current flagship, with better audio tags, cloning and 90+ languages than v3. It has no SSML and no Style or Speed sliders, only Stability and Similarity; use tags and punctuation for delivery.

Free. We don't store what you type.

How to prompt ElevenLabs

  1. Write only the words to be spoken, and steer delivery with inline [square-bracket] audio tags placed just before the words they affect.
  2. Use emotion and delivery tags such as [excited], [curious], [sarcastic], [whispers], [mischievously], [crying].
  3. Use non-verbal tags such as [laughs], [laughs harder], [starts laughing], [sighs], [exhales], [snorts], [gulps], [wheezing].
  4. Add sound-effect or accent tags where needed: [applause], [gunshot], [explosion], [strong British accent], [sings].
  5. Pace with punctuation: ellipses (…) for pauses and weight, CAPS for emphasis.
  6. Pick a voice whose natural style matches the tags you'll use; off-style tags work less reliably.
  7. Use tags only for things you can hear, not visual stage directions like [standing] or [grinning].
  8. For multi-speaker dialogue, put each speaker on their own line and assign them distinct voices.
  9. Don't change the script wording to fit tags in, and don't stack contradictory tags.
  10. For sound effects, describe the source, material, action and space concretely, with sequences joined by 'then'.
  11. Use audio terms for SFX: impact, whoosh, ambience, one-shot, loop, braam, glitch, drone; add BPM for musical loops.

The shape of a good prompt

TTS:
[emotion/delivery tag] [Line one]… [non-verbal tag] [Line two with KEY word emphasized].
(Dialogue: Speaker A: [tag] line / Speaker B: [tag] line)
SFX:
[Sound source + material + action], [environment/space], [character: one-shot/loop/impact], [then: follow-up sound]

Avoid

  • SSML <break> tags (not supported in v4)
  • Visual or stage-direction tags ([pacing], [smiling])
  • Narration like 'she said excitedly' (it gets spoken aloud)
  • Contradictory tags on the same line
  • Vague SFX prompts ('cool sound')

Example

Request an excited announcement that the cafe is reopening

[excited] Guess what? After three long months… we're BACK! [laughs] The doors open this Saturday at eight. [whispers] And the first coffee… is on us. [cheerfully] See you there!

The Weekly Tally

Get the latest AI tool updates, insights and no-hype advice.

A short, useful newsletter for people who actually use AI tools. Read this week's digest.

One email a week. Unsubscribe in one click.