Kubeez LogoKubeez
    School
    All lessons

    Make expressive text to speech with audio tags

    Type [whispering] before a line and Gemini 3.1 Flash TTS whispers it. This lesson shows which tags work, where to put them, and how Style instructions set the whole performance, with real takes you can play with and without tags.

    Audio6 min7 Steps

    What you will make

    A voiceover that whispers, sighs, laughs and pauses exactly where you want it to, in any language Gemini 3.1 Flash TTS speaks.

    You need

    • 1.The words you want spoken
    • 2.An idea of how they should sound
    • 3.Credits for a few takes

    1. 1. Open Text to Speech on the Gemini 3.1 Flash TTS tab

      Audio tags only work with Gemini 3.1 Flash TTS. The other voice tabs on the page do not perform them, so a tag there changes nothing.

      1. 1Open Text to Speech from the sidebar.
      2. 2At the top of the panel, pick the Gemini 3.1 Flash TTS tab, the one with the mark shown here.
      3. 3Pick a voice from the library and set Language to the language you will write in.

      Tip: The page can open on another tab. Check the tab before you type, or your tags will not be performed.

      Supports markup tags like [sigh], [laughing], [whispering], [shouting], [extremely fast], or even [like dracula] for expressive delivery.

      Gemini 3.1 Flash TTS is the tab with this mark. It is the only one that performs tags.

    2. 2. Write the line, then put a tag right before it

      The text box is read word for word. A tag in square brackets is not spoken. It changes how the words after it are said.

      1. 1Type the words exactly as they should sound.
      2. 2Put [whispering] at the start, before the first word.
      3. 3Generate, then play it next to the same line without the tag.
      Prompt
      [whispering] Go there. Quietly, before anyone sees you.

      Tip: A tag at the start sets the whole line. To change the delivery halfway, put a second tag where the change happens, like [whispering] then [shouting].

      Without tagsKore

      Go there. Quietly, before anyone sees you.

      With tagsKore

      [whispering] Go there. Quietly, before anyone sees you.

    3. 3. Add sounds: sighs, laughs and hesitations

      Some tags become a sound instead of a change in voice. [sigh] adds a sigh, [laughing] a laugh, [uhm] a hesitation and [gasp] a gasp. The tag itself is never read out.

      1. 1Put the sound tag exactly where the sound should happen.
      2. 2Keep real words after it, so the voice has something to say.
      3. 3Describe the laugh if you need a certain kind, like [laughing softly] or [laughing hard].
      Prompt
      [sigh] Fine. [short pause] I'll fix it again tomorrow.

      Tip: The kind of sigh or laugh follows the words around it. Funny words give an amused laugh, tired words a tired sigh.

      Without tagsSulafat

      Fine. I'll fix it again tomorrow.

      With tagsSulafat

      [sigh] Fine. [short pause] I'll fix it again tomorrow.

      Without tagsPuck

      I can't believe you actually did that. That's the funniest thing I've ever seen.

      With tagsPuck

      [laughing] I can't believe you actually did that. [laughing hard] That's the funniest thing I've ever seen.

    4. 4. Control the timing with pause tags

      Pause tags hold the silence for you. [short pause] is about a quarter of a second, [medium pause] about half a second, and [long pause] a second or more.

      1. 1Put [long pause] right before a reveal or a punchline.
      2. 2Use [short pause] for a breath between two thoughts.
      3. 3Let commas and three dots handle the small, natural pauses.
      Prompt
      And the winner is... [long pause] Maria.

      Tip: Use pauses where they mean something. A pause after every sentence sounds slow and unnatural.

      Without tagsCharon

      And the winner is... Maria.

      With tagsCharon

      And the winner is... [long pause] Maria.

    5. 5. Mix a few tags in one line

      One line can carry several tags, one for each moment. A few well placed tags sound better than a tag on every word.

      1. 1Start with a tag for the mood of the line, like [nervously].
      2. 2Add a sound tag where the moment happens, like [uhm] or [gasp].
      3. 3Keep words or punctuation between two tags. Never put two tags side by side.
      Prompt
      [nervously] Um, hi. [uhm] I think I sent that email to the whole company. [gasp] Oh no.

      Tip: Write the words the way people really talk, with the "um" and the "oh no". Tags work better on words that already carry the feeling.

      Without tagsZephyr

      Um, hi. I think I sent that email to the whole company. Oh no.

      With tagsZephyr

      [nervously] Um, hi. [uhm] I think I sent that email to the whole company. [gasp] Oh no.

    6. 6. Set the whole performance in Style instructions

      Tags shape moments. The Style instructions field under the text box sets the whole take: who is speaking, the mood, the pace and the accent. Use both together.

      1. 1Describe the character and how they speak, in plain words.
      2. 2Keep the tags in the text for the moments inside the line.
      3. 3Play the three takes here: no tags, tags only, then tags with Style instructions.
      Prompt
      Speak like a theatrical vampire host. Slow, velvety, menacing, savoring every word.

      Tip: Be specific. "British English accent as heard in Croydon, England" works better than "British accent", and a character works better than a single feeling.

      Without tagsAlgenib

      Good evening. I have been... expecting you.

      With tagsAlgenib

      [whispering] Good evening. [long pause] I have been... [laughing softly] expecting you.

      Tags and Style instructionsAlgenib

      Speak like a theatrical vampire host. Slow, velvety, menacing, savoring every word.

      [whispering] Good evening. [long pause] I have been... [laughing softly] expecting you.

    7. 7. Use tags in other languages

      Tags work on words in any language Gemini 3.1 Flash TTS speaks. Write the words in your language and keep the tags in English.

      1. 1Set Language to the language of your text.
      2. 2Write the words in that language.
      3. 3Keep every tag in English, like [whispering], not a translated word.
      Prompt
      [whispering] Ve allí. En silencio, antes de que alguien te vea.

      Tip: Style instructions can stay in English too. They are directions for the voice, not words it says.

      EnglishKore

      [whispering] Go there. Quietly, before anyone sees you.

      SpanishKore

      [whispering] Ve allí. En silencio, antes de que alguien te vea.

      RomanianKore

      [whispering] Vino mai aproape. Îți spun un secret, dar să rămână între noi.

    Avoid these

    • Typing tags on another voice tab. Only Gemini 3.1 Flash TTS performs them.
    • Describing the sound instead of writing it. "She laughs softly" gets read out loud. Write the words and put [laughing softly] before them.
    • A line made only of tags, with no words to say.
    • Two tags side by side, like [sigh] [laughing]. Put words or punctuation between them.
    • A tag on every sentence. The take sounds overacted.
    • Translating the tags. Keep them in English, even when the words are not.
    • Trusting a new one word feeling tag without a test. Some, like [scared], [curious] or [bored], can be read out as a word. Put that feeling in Style instructions instead.

    Audio tags in Gemini 3.1 Flash TTS

    Short cues in square brackets, like [whispering] or [sigh], placed in the text right before the words they shape. The voice performs them instead of reading them, so you can make it whisper, laugh, sigh or pause at an exact spot.

    Put [whispering] before the words on the Gemini 3.1 Flash TTS tab of Text to Speech. For a take that stays in a whisper from start to end, also say so in Style instructions.

    No, there is no fixed list. Common ones are [whispering], [shouting], [laughing], [sigh], [gasp], [uhm], [sarcasm], [excited], [short pause], [medium pause] and [long pause]. Clear cues like [like dracula] or [extremely fast] also work, so test a new tag before you rely on it.

    Usually not. Sound, style and pause tags are performed, not spoken. Some one word feeling tags, like [scared], [curious] or [bored], can be read out as a word, so test them or put the feeling in Style instructions.

    Type [short pause], [medium pause] or [long pause] where the silence should go. They last about a quarter of a second, half a second, and a second or more. Commas and three dots give smaller pauses.

    Style instructions set the whole take, like the character, mood, pace and accent. Tags change a single moment inside the line. The best takes use both.

    Yes. Write the words in your language, set Language to match, and keep the tags in English.

    The safety check sometimes flags a harmless line. Generate it again, and if it keeps failing, change a few words.

    Your turn

    Everything in this lesson is one click away. Open the tool and make your first version now.

    Try it in Kubeez