ElevenLabs Text to Speech Emotion Tags: The Complete List

ElevenLabs text to speech emotion tags are bracketed cues like [whispering], [laughing], and [excited] that steer delivery on Eleven v3 and v4 models. You can type them manually or let the Enhance button add them automatically. This elevenlabs text to speech emotion tags guide lists them all. Checked 4 Nov 2026.

What are ElevenLabs emotion tags?

Emotion tags — officially called audio tags in ElevenLabs documentation — are bracketed cues you place inside your text to direct how the voice delivers it. According to ElevenLabs’ prompting guide for Eleven v4, tags like [whispering], [shouting], and [laughing] give you fine-grained control over delivery, and v4 handles them with more nuance than previous models. The concept is simple: instead of hoping the model guesses the emotion, you tell it explicitly.

Tags work on Eleven v3 and v4, the models built for expressive, director-style control. ElevenLabs notes that being explicit helps: tags land more readily when the delivery you request is represented in the voice’s training data. A dramatic stage whisper works better on a voice trained with expressive material than on a flat corporate narrator — voice selection still matters.

One important technical note from the docs: Eleven v4 and v3 do not support SSML break tags. For pauses and pacing, use audio tags, punctuation like ellipses, and text structure instead. If you are coming from SSML-based tools, this is the main syntax change to absorb. For more on prompting these models, read our ElevenLabs v3 prompts guide.

The complete emotion tag list

ElevenLabs documents its audio tags in the v4 prompting guide and the Enhance-prompt reference. Here are the tags, grouped by what they do:

A few usage notes from the official guide: tags go in square brackets directly in your text, and the Enhance button in the ElevenLabs UI can auto-generate audio tags for you. The [strong X accent] tag is a template — replace X with the accent you want, such as [strong British accent]. And remember that explicit direction yields more predictable results than subtle hints, so write tags that say exactly what you mean.

How to use emotion tags in text to speech

Follow these steps, with what each one costs in credits:

1. Select an Eleven v3 or v4 model. Audio tags need the expressive models — v3 (70+ languages, 5,000-character limit) or eleven_v4 (90+ languages, 10,000-character limit). Older models ignore the tags. Cost: 0 credits to select.
2. Write your script with tags in brackets. Place cues like [excited] or [whispering] directly before the lines they affect. Keep tags explicit — “[angry]” beats a vague hint. Cost: 0 credits to write; text to speech costs one credit per character when you generate.
3. Use Enhance to auto-generate tags (optional). In the ElevenLabs UI, the Enhance button can add audio tags for you automatically — useful when you have a long script and want a first pass of direction. Cost: 0 credits to enhance.
4. Generate the audio. Run text to speech and listen to how each tagged section lands. Remember that tags work best when the delivery exists in the voice’s training data, so a mismatch may need a different voice. Cost: one credit per character generated.
5. Refine and regenerate. Adjust tags that landed weakly — make them more explicit or reposition them — and regenerate. Short test generations cost fewer credits than re-running a full chapter. Cost: one credit per character per regeneration.

Tips for emotion tags that actually land

Match the tag to the voice. ElevenLabs’ guide stresses that tags land more readily when the delivery is in the voice’s training data, so audition voices with the emotion you need most. A voice built for audiobook narration will sell [thoughtful] and [sad] better than [shouting], while a character voice may do the reverse.

Be explicit and specific. The docs say that being explicit helps — clear dialogue tags yield more predictable results than subtle ones. Write [furious] rather than hoping [annoyed] escalates, and put the tag immediately before the line it governs so the model binds them together.

Layer tags with punctuation. Since v3 and v4 do not support SSML break tags, combine audio tags with ellipses for hesitant pacing and standard punctuation for rhythm. A line like “[whispering] I don’t think we should be here…” uses the tag for tone and the ellipsis for the trailing, uncertain pause — each doing a different job.

Test tags on short samples first. Before running a full chapter or long script, generate a 2–3 sentence sample with your key tags to hear how the voice handles them. This costs only a handful of credits at one credit per character, and it reveals mismatches — a [shouting] tag that sounds strained, or a [whispering] tag the voice can’t sustain — before you commit to a long render. Adjust the voice or the tags based on the sample, then run the full text with confidence.

On cost, text to speech runs one credit per character, and plans scale from Free $0 (10,000 credits/mo) to Starter $6/mo (30,000), Creator $22/mo (121,000), and Pro $99/mo (600,000) — each price checked 4 Nov 2026. A $1 Starter first-month offer was shown until Nov 8 when checked on 4 Nov 2026. Because tags themselves add characters, a heavily tagged script costs slightly more than plain text — keep tags purposeful rather than decorative.

Who should NOT use it

Skip emotion tags if you are generating on older ElevenLabs models — the tags will be ignored and you will pay per character for bracketed text that does nothing. They are also overkill for flat, informational narration like terms-and-conditions reads, where neutral delivery is the point. And if you need sample-exact pause durations for broadcast compliance, the lack of SSML break support on v3/v4 makes tags the wrong precision tool.

Frequently asked questions

What are ElevenLabs text to speech emotion tags?

They are bracketed audio tags like [whispering], [laughing], and [excited] that direct vocal delivery on Eleven v3 and v4 models. According to ElevenLabs’ prompting guide, the tags give fine-grained control — you tell the model exactly how to deliver each line instead of hoping it guesses.

How do I add emotion tags to my script?

Type the tag in square brackets directly before the line it affects, for example “[excited] We did it!” Alternatively, use the Enhance button in the ElevenLabs UI to auto-generate audio tags across your script. Both approaches work on Eleven v3 and v4.

Which ElevenLabs models support emotion tags?

Eleven v3 and Eleven v4. The v4 prompting guide documents audio tags with the most nuance, while v3 also supports them across 70+ languages. Older models do not support audio tags, so select v3 or v4 in the model picker before generating.

Can emotion tags do accents?

Yes, with the [strong X accent] template tag — replace X with the accent you want, such as [strong British accent]. ElevenLabs documents this as a special tag. Results vary by voice, since the accent needs to be plausible for the voice’s training data.

Do emotion tags work with SSML?

No — Eleven v4 and v3 do not support SSML break tags, according to the official docs. Use audio tags, punctuation like ellipses, and text structure for pacing instead. If your workflow depends on SSML, check model compatibility before committing.

Do emotion tags cost extra credits?

Tags themselves have no separate fee, but text to speech costs one credit per character — and tag characters count. A heavily tagged script therefore costs slightly more than the same script without tags. Keep tags purposeful: every [excited] should earn its characters.

Sources checked

• ElevenLabs documentation: Prompting Eleven v4 / audio tags guide — https://elevenlabs.io/docs/overview/capabilities/text-to-speech/best-practices — checked 4 Nov 2026
• ElevenLabs documentation home (voice IDs, credit costs, model specs) — https://elevenlabs.io/docs/overview/intro — checked 4 Nov 2026