Emotion Tags in AI Text to Speech: How to Direct a Voice
Updated September 30, 2026
TL;DR: on LeapFun's Leap Ultra and Leap Pro multilingual voices you direct a read two ways: inline tags such as [excited] or [whispers] placed before the words they affect, and a director's note for the whole take (style, pace, accent). Turbo, Expressive and the Chinese dialect voices read tags as literal text, so the app strips them when you switch. Tags change delivery, not the script.
The tags Ultra and Pro voices understand
Type [ in the script box and the app suggests them: 22 in all, including [laughing], [sighs] and pause marks. The core set:
| Tag | What it does | Use it for |
|---|---|---|
| [excited] | Lifts energy and pace | Hooks, reveals, offers |
| [happy] | Warm, smiling delivery | Friendly explainers, UGC |
| [calm] | Slower, settled | Meditation, sensitive topics, closing lines |
| [serious] | Level, weighty | Documentary, warnings, finance |
| [gentle] | Soft and unhurried | Bedtime stories, apologies |
| [sad] | Lowered, slower | Emotional beats in fiction |
| [angry] | Sharper attack | Character dialogue |
| [surprised] | Raised, quicker | Reactions, plot turns |
| [embarrassed] | Hesitant, softer | Comedy, confessions |
| [whispers] | Whispered | Intimacy, ASMR-style hooks |
| [breathy] | Airy texture | Trailers, romance |
| [emphasis] | Stress on what follows | Key numbers, names, the CTA |
Where to put them
Directly before the sentence or clause you want changed. A tag carries until the mood naturally shifts or the next tag. One tag per beat is plenty; stacking three in a sentence produces a confused read.
[serious] In 1977, two spacecraft left Earth carrying a golden record. [calm] Nearly fifty years later, they're still going. [excited] Here's how far they've gone.
The director's note
For the whole take, open Advanced and set style, pace and accent from the presets, or write a note in plain language: "warm documentary narrator, unhurried, slight smile, British accent." Use the note for the baseline and tags for the exceptions. This is the same mechanism a session director uses: the note is the brief, the tags are the pencil marks on the script.
What the presets cover
| Control | Presets | Use it for |
|---|---|---|
| Style | Broadcast news, vocal smile, whisper, calm storytelling, high-energy promo, deadpan | The overall character of the take |
| Pace | Natural, rapid-fire, slow and unhurried, staccato | Matching the cut: a 15-second ad versus a meditation |
| Accent | American, British, Australian English | Regional fit without changing the voice |
| Free-text note | Up to 300 characters in your own words | Anything the presets don't say |
Hear it: one script, two notes
Same voice, same lines — “Ninety seconds. That's all it took for the whole plan to fall apart. And nobody in the room said a word.” — with only the note changed.
Previews above are real LeapFun output. Hit Generate to hear your script — 3,000 free credits on sign-up, no card required.
The note works on Leap Ultra and on Leap Pro's multilingual voices, which are the same 30 voices at the standard rate, and it applies in any of the 130+ languages those voices read: write the note in English and the script in the language you are publishing in.
What happens on other voice tiers
Leap Turbo and Expressive voices, and Pro's Mandarin and dialect voices, don't interpret tags ("Expressive" names that tier's natural delivery, not tag control) — they'd read "excited" aloud. So when you switch from a multilingual voice to one of those, LeapFun strips the tags from the script and tells you. Emotion on those tiers comes from the voice choice and from punctuation: a full stop is a breath, a comma is a pause, an ellipsis is a hang.
Try it
Paste the example above on the text-to-speech page with an Ultra voice — Stella or Leo — and compare with and without tags.
Frequently asked questions
Do emotion tags cost extra credits?
Only by a few characters. Tags are part of the text you send, so a tag like [whispers] is billed like any other characters at the voice's normal rate — a handful of credits, negligible next to the script itself.
Why did my tags disappear?
You switched to a voice that doesn't interpret tags: a Turbo or Expressive voice, or one of Pro's dialect voices. Those would read the tags aloud, so the app removes them and shows a notice. Switch back to an Ultra or Pro multilingual voice to use them.
Can I make a voice whisper a whole paragraph?
Yes — put [whispers] at the start; it carries until the next tag or a natural shift. For a whole take, use the director's note instead.
Do tags work in Audiobook Studio?
Yes, paragraph by paragraph on Ultra and Pro multilingual voices; new Studio projects start on an Ultra narrator. Combine with character voice assignments for dialogue.
Try it on your script
3,000 free credits on sign-up, no card — generate a first take in seconds.
Start free See pricing