Thriller
[low, urgent] on the phone call that changes everything, and [breathe] before the door opens. More in thriller audiobook narration.
Directions and sound tags
AI Voice Emotion Tags That Tell a Line How to Feel
AI voice emotion tags are short notes in square brackets. Put one at the start of a line, such as [whisper] or [like a late-night radio host], and the voice changes how it reads; drop a sound tag such as [laugh] or [sigh] inside a sentence and it happens right there. The tool above opens on the Expressive model, which follows them most closely.
AI audiobook samples
View allFour directed lines, played back word by word
Each card pairs a written direction with inline sound tags. Press play and follow the words as they are spoken.
Nate
excited delivery
[say excitedly, like sharing big news with a friend]
Guess what? We actually did it. [laugh] The whole thing launches tomorrow morning!
Luna
whisper delivery
[in a hushed, breathy whisper, like sharing a secret in the dark]
Come closer. [breathe] I'll tell you where the old lighthouse keeper hid the key.
Rupert
anchor delivery
[say like a confident, authoritative news anchor, crisp and measured, stressing the key words]
Markets opened higher this morning, as investors welcomed news of a long-awaited trade agreement.
Nadia
ad delivery
[say warmly and cheerfully like an upbeat radio ad, smiling through every word]
Fresh bread, real coffee, and a table by the window. Your morning starts at Rosie's Corner Bakery.
Directions and sound tags: the two kinds of emotion tags
Both are typed in square brackets. Where you put the tag decides what it does.
| Directions | Sound tags | |
|---|---|---|
| Where it goes | At the start of a line | Anywhere inside the line |
| What it changes | The delivery of the whole line | One moment: a laugh, a sigh, a breath |
| Examples | [whisper], [furious], [trembling, holding back tears] | [laugh], [sigh], [breathe], [whisper] |
| Standard, Lite and Expressive | Performed | Performed |
| Premium (Turbo 2.5) | Ignored | Removed before reading, never spoken |
How to add AI voice emotion tags to a chapter
- 1
Mark the key moments
Read the chapter once and note the reveals, the arguments and the last line. Those are the only places most chapters need a tag.
- 2
Write the direction
Type a note in brackets at the start of the line, or use the direction button on any line in the studio. Keep it to a few plain words.
- 3
Add a sound if it helps
Place [laugh], [sigh] or [breathe] exactly where it would happen in the sentence, not at the end by habit.
- 4
Regenerate that line
Play the line back. If it overshoots, soften the note and regenerate only that line, leaving every other line untouched.
A starter list of emotion tags for narration
There is no fixed list: any short, plain description works. These are reliable starting points for long-form audio.
| Tag | Use it for | Tip |
|---|---|---|
| [whisper] | Secrets, night scenes, menace up close | Follow it with a normal line so the drop lands |
| [furious, through clenched teeth] | Confrontations and ultimatums | Save it for the peak of the scene |
| [trembling, holding back tears] | Grief, confessions, goodbyes | Works best on short lines |
| [warm, smiling] | Romance, reunions, a kind narrator | Subtle enough to use more often |
| [dry, sarcastic] | Banter and comic asides | Pairs well with a [sigh] |
| [laugh] | Laughter in the middle of a line | Expressive voices do it most naturally |
| [sigh] | Weariness or relief | Put it where the breath would fall |
| [breathe] | A beat before a reveal | Once or twice a chapter is plenty |
Preparing a manuscript with tags in place? The audiobook script format guide shows a clean layout.
[say with hyped, high energy like a YouTube intro]
Stop scrolling. In the next sixty seconds, I'm going to show you the trick that doubled my channel.
Writing directions an AI voice with emotion will follow
Good directions are short and specific. [nervous] works; [quietly, like she is afraid of being overheard] works better, because it gives the voice a reason. Comparisons help too: [like a late-night radio host] or [like a teacher who has said this ten times] pack tone, pace and attitude into a few words.
Keep the note consistent with the text. A line that says “she shouted” and a [whisper] tag pull in opposite directions, and the result sounds confused. If a character has a signature laugh or sigh, use the same sound tag each time so it becomes part of the voice.
Where emotion tags earn their place, genre by genre
Romance
[warm, teasing] for banter, [barely above a whisper] for the confession. One well-placed [laugh] does more than ten.
Horror
[slow, hushed dread] for the narrator as the scene tightens, and a sudden [whisper] from somewhere it should not be. See horror audiobook narration.
Fantasy
[commanding, echoing] for the queen’s decree, [gravelly, amused] for the old mentor who has seen it all.
Children’s
[giggling] for the little rabbit, [big and booming] for the friendly giant. Children love exaggeration, so tags can be bolder.
Nonfiction
Rarely needed. A [warm] or [conversational] note on the opening paragraph sets the tone for a whole chapter.
When sound tags help and when they hurt
Tags are seasoning. A chapter with a tag on every line sounds acted from start to finish.
Use tags for
- The climax of a scene, where flat delivery would undercut the writing
- A reveal or a twist, with a [breathe] just before it
- The last line of a chapter, which listeners carry into the next one
- A character’s signature laugh or sigh, repeated across the book
Skip tags for
- Every line of dialogue in a scene
- Plain descriptive narration, which reads best neutral
- Stacking three or more tags on one short line
- Lines where the tag contradicts the text around it
Which models follow AI voice emotion tags
Expressive (ElevenLabs v3). The widest emotional range: real laughter, whispers that sound close to the microphone and voices that break on a sad line. It costs 34 credits per 1,000 characters, so many authors save it for dialogue-heavy chapters or the key scenes of a multi-voice audiobook. It shares its 67 voices with Premium.
Standard and Lite (Gemini Flash TTS family). Both follow directions at the start of a line and inline tags such as [laugh] and [sigh], at 9 and 6 credits per 1,000 characters. For most AI audiobooks, Standard with a handful of directions per chapter is the best balance of cost and feeling.
Premium (ElevenLabs Turbo 2.5). Steady and fast, with 67 distinct character voices, but it does not perform tags. Brackets are stripped before the text is read, so a stray [laugh] is never spoken aloud. Switch the project to Expressive to keep the same voices and direct them.
Delivery and cost. A project setting, from stable through balanced to creative, controls how much a voice varies between takes; creative gives tags more room. Tags themselves are only a few characters, so they barely change the cost of a chapter. Browse the voices that take tags best on the AI voices for audiobooks page.
Ai voice emotion tags FAQ
What are AI voice emotion tags?
They are short notes in square brackets that tell a text to speech voice how to perform a line. A direction at the start, such as [furious], shapes the whole line, while a sound tag inside it, such as [laugh], adds a single moment of laughter, a sigh or a breath.
Is there a fixed list of tags I have to use?
No. Directions are plain English, so [like a tired detective at 3am] works as well as [tired]. Sound tags such as [laugh], [sigh], [breathe] and [whisper] are the most reliable inline sounds. If a note does nothing, rephrase it more simply.
Will a tag ever be read out loud as a word?
On Premium voices, no: brackets are removed before reading. On Standard, Lite and Expressive the tag is performed rather than spoken. If an unusual tag comes out oddly, simplify the wording and regenerate just that line.
Does adding tags change the price of a chapter?
Barely. Credits are charged per 1,000 characters and a tag is only a few characters long. The bigger cost decision is the model: Expressive, which follows emotion tags most closely, uses 34 credits per 1,000 characters against 9 for Standard.
How many tags should a chapter have?
Usually five to ten, placed on the turning points: arguments, reveals, confessions and the final line. Neutral narration between them makes the tagged moments stand out. If every line has a tag, none of them land.
Why did my tag seem to do nothing?
The project may be on the Premium model, which ignores tags, or the note may be too subtle for the voice. Switch to Expressive, try a stronger or more specific description, and regenerate the line a second time, since each take varies a little.
Try AI voice emotion tags on a scene of your own
Paste a few lines, add [whisper] or [laugh] where it matters and hear the Expressive model perform them with 100 gift credits (valid for 7 days, about 12 minutes on Standard; usage varies by text and voice tier) from a Google sign-up.