Emotional Text to Speech: Audio Tags with Samples
Emotional text to speech on Eleven v4: write [crying], [whispers] or [angry] before a line and the voice acts it. Tested tags, A/B samples and prices.
Last updated: 2026-10-26
Emotional text to speech is AI speech that performs a feeling, a sob, a whisper, a laugh, a raised voice, instead of reading every line in the same even tone. On Eleven v4 you direct it with an audio tag in square brackets right before the words, as in [crying] I kept your letter., and the tag is acted, never read aloud. Every claim on this page comes with audio: one line in four deliveries, tags against long written directions, and lines in Mandarin and Japanese.
Hear the voices used on this page
Tap a voice to hear it. To voice your own text, paste it into the workbench: 7 credits per 100 characters, starting with the $3.90 trial pack.
One line, four deliveries
Same voice, same words, four separate Eleven v4 takes. The first has no tag; then [crying], [whispers] and [shouting]:
[crying] I kept your letter. I read it every night.
[whispers] I kept your letter. I read it every night.
[shouting] I kept your letter. I read it every night.The sentence carries no emotion of its own, so whatever changes from take to take comes from the tag. That is the whole trick of emotional TTS on Eleven v4: write the words plainly and let the tag direct them.
Where the emotion comes from
| Lever | What it controls | Example |
|---|---|---|
| The model | Whether tags are performed at all | Eleven v4 and v3 act them; remove tags before switching to Multilingual v2 or Flash |
| Audio tags | The feeling, and sounds like sobs, sighs and laughs | [crying], [whispers], [laughs] |
| The words | What the feeling is about; ElevenLabs says v4 reads the scene, not each sentence alone | "I kept your letter." |
| Punctuation | Pace and hesitation | ..., dashes, short sentences, line breaks |
| The voice | Range and timbre | A bright voice laughs easily; a low one menaces |
Emotion tags that work, by emotion
These are tags we have used on Eleven v4 in our own tests and finished episodes. Each one has a clip further down this page.
| Emotion | Tags to write | Where to hear it |
|---|---|---|
| Sad, crying | [crying], [sobbing] | One line, four deliveries; Choking up |
| Angry | [angry], [menacing], [shouting] | Holding back anger; One line, four deliveries |
| Afraid | [fearful], [nervous] | Fear; Change emotion inside one line |
| Whispering | [whispers] | One line, four deliveries; the Mandarin sample |
| Laughing | [laughs] | Change emotion inside one line; Mandarin and Japanese samples |
| Sighing | [sighs] | Change emotion inside one line; the Mandarin sample |
| Excited | [excited] | The Japanese sample |
| Sarcastic | [sarcastic] | The Japanese sample |
| Calm, serious | [calmly], [solemnly] | The butler sample; Fear |
Our audio drama studio directs every scene from a set of 27 such tags. The longer list, with tips on placement, is in the emotion tags cheat sheet.
Tags or written-out direction?
Eleven v4 also accepts direction in your own words, such as [barely holding it together, forcing a smile]. We tested both on the demo scripts Fish Audio published for its Drama 3 model: each script rendered once exactly as written, with long free-form directions, and once with standard v4 tags, in the same voice, one take each, then transcribed with ElevenLabs Scribe. What we found:
- No direction was ever spoken. None of the 11 Eleven v4 takes read a direction aloud, including one written in Chinese and one with
<emphasis>markup. - Some free-form directions were acted. "Cracking up despite herself" produced a real laugh.
- One slip. The verbatim take of the anger script said "Just one" twice. The tagged take did not. With one take each, we can't say how often that happens.
Choking up
Written direction, [choked up, crying quietly] … [voice breaking] …:
[voice breaking] I keep thinking he'll walk in and ask for it.Tags, [crying] … [sobbing] …:
[crying] I left his jacket by the door. [sobbing] I keep thinking he'll walk in and ask for it.Holding back anger
Written direction, [through gritted teeth, keeping his voice down] … [slow, dangerously quiet] …. This is the take that repeated "Just one":
Tags, [angry] … [menacing] …:
[angry] I asked you to do one thing tonight. [menacing] Just one.Fear
Written direction, [breathless, speaking fast] … [voice drops to a terrified whisper] … [slow, deadly serious] …:
[voice drops to a terrified whisper] And whatever you hear out there… [slow, deadly serious] <emphasis>do not open that door.</emphasis>Tags, [fearful] … [whispers] … [solemnly] …:
[fearful] Just turn off the lights. Stay away from the windows. [whispers] And whatever you hear out there… [solemnly] do not open that door.When to use which. Use a standard tag when one emotion covers the line and you want much the same result every time. Write the direction out when the feeling is mixed, like "forcing a smile" or "laughing despite herself", which no single tag says; expect more variation, so generate twice and keep the better take. All 11 takes and Fish's own versions are on Fish Audio Drama 3 vs Eleven v4.
The voices in these A/B clips (Alice, Brian, Laura) are ElevenLabs default voices that retire on December 31, 2026; see ElevenLabs default voices. The other samples on this page use voices that are not retiring.
Change emotion inside one line
Eleven v4 follows tags in the order you write them, so one generation can move from calm to panic. Put a tag at each turn in the feeling, not before every sentence:
[whispers] Okay, don't move. I think it's still in the kitchen. [nervous] What if it heard us? [laughs] Relax, it's just the cat. [sighs] You scared me half to death.A dry butler losing his composure, four tags in one take:
[calmly] The guests have arrived, sir. [nervous] All of them. [whispers] Including the ones we buried last spring. [sighs] I shall set two more places.Two tags can also sit side by side when a moment needs both, as in [chuckles] [whispers] Go on without me.
Emotional text to speech in Chinese and Japanese
Keep the tag in English and write the line in any language Eleven v4 speaks. Mandarin, with [laughs], [whispers] and [sighs]:
[laughs] 你以为我会就这么算了？[whispers] 合同在我手里，三天之内，你会求着来找我。[sighs] 可惜啊，我们本来可以是朋友。Japanese, two voices, with [excited], [sarcastic], [laughs] and [whispers]:
[excited] みなさーん、こんばんは！今夜の目玉はこちら、ふわふわクラウドのノイズキャンセリングヘッドホン！
[sarcastic] バッテリーは四十時間。お正月の親戚の長話も、まるごと聞き流せます。
[laughs] 通常は一万八千円のところ、今夜だけ、なんと九千八百円！
[whispers] 九千八百円……二回確認しました。正直、ちょっと驚いています。
[excited] 在庫は残り二百台！カウントダウン、いきますよ！さん、にー、いち！
リンク、出しました。お早めにどうぞ。Pauses, pacing and emphasis
- Ellipses slow a line down and add hesitation: "I kept it… every night."
- Short sentences and line breaks make tension; long ones relax it.
- A dash cuts a thought off mid-word, for interruptions.
- Capitals on a single word add emphasis. Use them sparingly.
- No SSML needed. Eleven v4 and v3 take tags and punctuation instead. For names and jargon that come out wrong, use the pronunciation dictionary.
Voices that carry emotion well
On Eleven v4 the model does most of the acting, but a voice's natural range still matters:
| Voice | Designed for | What it does well | Sample on this page |
|---|---|---|---|
| Sweet Girl (EN) | English | Whispers, nerves, laughter, tears | One line, four deliveries |
| British Butler | British English | Dry calm that cracks | The butler sample |
| Ice Queen | Mandarin | Cold laughter, threats in a whisper | The Mandarin sample |
| Anime Girl | Japanese | Bright excitement, quick laughs | The Japanese sample |
For new work, avoid ElevenLabs' stock default voices such as Sarah, Brian, Lily and Jessica: all of them retire on December 31, 2026. Our designed character voices are not on that list; hear all 24 in the voice library.
A whole emotional scene
For a scene with several characters, let the audio drama studio do the directing: it reads the story, casts a voice for each character and tags the lines where the text implies a feeling. You can change any line's emotion before rendering. This episode was made that way, crying, whispers and a cold reply included:
To pick every voice and tag yourself, use multi-speaker text to speech, which renders a whole scene in one generation. For how v4 compares with the model before it, see Eleven v4 vs v3.
What emotional text to speech costs
Emotion costs nothing extra: 7 credits per 100 characters on Eleven v4, the same as plain speech. In the text-to-speech tool the tags count as characters ([whispers] is 10); in the audio drama studio, the tag at the start of each line is not counted.
- 1,000 characters, about 70 seconds of English, cost 70 credits: about $1.46–2.32 at monthly-plan list prices ($0.58–0.92 in the first month at the new-user price).
- The trial pack has 150 credits, about 2,100 characters, one-time and never expiring: $9.90 list, or $3.90 for new users in the first 24 hours after sign-up.
- Nothing is free to generate, but every sample on this page, every voice preview and the 534 episodes in the theater are free to listen to. See pricing.
FAQ
How do I make text to speech sound emotional? Use a model that performs audio tags, Eleven v4 or v3, and write the emotion in square brackets right before the words: [crying], [whispers], [angry], [laughs]. Punctuation does the rest: ellipses for hesitation, short sentences for tension. Tags belong with Eleven v4 and v3; remove them before switching to Multilingual v2 or Flash.
Does the AI read the tags out loud? Not in our tests. In our Eleven v4 vs v3 test, no tag was read aloud on either model, and in 11 Eleven v4 takes written with long free-form directions, no direction turned up in the speech-to-text transcripts. Tags never appear in the subtitles either.
Can text to speech cry, whisper or shout? Yes. On Eleven v4, [crying], [whispers] and [shouting] are all performed. This page has all three on the same line in the same voice, next to the line with no tag at all.
Which AI voice is best for emotional text to speech? The model matters more than the voice: on Eleven v4 most voices can cry, whisper and laugh. Pick a voice whose natural range suits the part, and for new projects avoid ElevenLabs' default voices: Sarah, Brian, Lily and 16 others retire on December 31, 2026.
Do emotion tags work in other languages? Yes. Write the tag in English and the line in any language Eleven v4 speaks; the Mandarin and Japanese samples on this page do exactly that.
How much does emotional text to speech cost? The same as plain speech: 7 credits per 100 characters on Eleven v4, and in the text-to-speech tool the tags count as characters. Listening to every sample is free; generating is not. The one-time trial pack has 150 credits (about 2,100 characters) at a $3.90 new-user price in the first 24 hours after sign-up (list $9.90).