What "sighs" means in ElevenLabs and why you'd add them

In ElevenLabs, sighs are breathing sounds — exhales, inhales, pauses where a person naturally catches their breath — that you layer into AI-generated speech to make it sound more human. The software generates the words, but without these sounds, the voice reads in a flat, robotic way. Adding sighs breaks up the monotony and signals emotion: frustration, relief, tiredness, or just the natural rhythm of how people actually talk.

You add sighs by inserting special text markers into your script before you generate the audio. ElevenLabs recognizes these markers and converts them into actual breathing sounds during the voice generation process. The markers are straightforward — you type them directly into your text — and they work across all ElevenLabs voices.

Key Takeaways

  • ElevenLabs uses text markers like [sighs], [breathes], and [clears throat] that you type into your script to add breathing sounds to generated speech.
  • These markers must be placed exactly where you want the sound to occur, usually at the end of a sentence or before a long pause in dialogue.
  • The markers work in the text input box on the ElevenLabs website or in the API if you are using the software programmatically.
  • Not all voices respond to markers equally — some render them more naturally than others, so test with your chosen voice first.

Where to find and use the markers in ElevenLabs

Open the ElevenLabs text-to-speech tool on the website (elevenlabs.io). In the text input box where you normally type the words you want spoken, place your cursor where you want a sigh to occur. Type the marker exactly as written: [sighs]. You can also use [breathes], [clears throat], [coughs], or [laughs] — these are all recognized by the system.

For example, if your script reads "I've been waiting all day for this," you could write it as "I've been waiting all day for this [sighs]" to add a sigh at the end. Or place it mid-sentence: "I've been [sighs] waiting all day for this" if you want the sigh to interrupt the speech. The position matters — ElevenLabs will generate the sound at that exact point in the audio.

After you type your script with the markers, select your voice from the dropdown menu and click the generate button. The audio will include the breathing sounds at the locations you marked. Listen to the result — if the sigh sounds unnatural or too loud, adjust its placement or remove it and regenerate.

Common markers and what they sound like

ElevenLabs recognizes several breathing and vocal markers. [sighs] produces a clear exhale, usually sounding like resignation or relief. [breathes] is a lighter inhale or exhale, useful for breaking up long sentences without adding emotion. [clears throat] adds a throat-clearing sound, good for transitions or when a character is about to speak seriously. [coughs] produces a cough, and [laughs] adds laughter.

The exact sound depends on the voice you choose. Some voices render these markers more naturally than others — a voice designed for narration might produce a subtle sigh, while a voice designed for character dialogue might produce something more exaggerated. Test your chosen voice with a marker before you commit to a full script. Generate a short sentence with [sighs] at the end, listen, and decide if it fits the tone you want.

Why markers sometimes don't work as expected

If you type a marker and the generated audio does not include the breathing sound, the most common reason is a typo. The markers are case-sensitive and must be spelled exactly right, with square brackets and no extra spaces. [sighs] works; [Sighs], [ sighs ], or [sigh] (without the 's') will not.

Another reason is that some voices are less responsive to markers than others. Voices trained on specific types of speech — like a news anchor or a calm narrator — may ignore markers or render them very subtly. If a marker produces no sound at all, try switching to a different voice and regenerating. You can also try placing the marker in a different position in the sentence, as some voices respond better to markers at sentence boundaries than in the middle of speech.

Using markers in longer scripts and conversations

If you are generating a longer piece with multiple speakers or a monologue with natural pauses, place markers at points where a real person would naturally breathe. After a long sentence, before a topic change, or when a character is thinking — these are all good places for a sigh or breath. Overusing markers makes the audio sound artificial, so use them sparingly.

For dialogue between characters, you can add markers to each speaker's lines separately. If you are generating one character at a time, add the marker to that character's script. If you are pasting a full conversation with speaker labels, place markers within each character's dialogue block. ElevenLabs will explore the marker only to the text when ready around it.

Testing and adjusting your markers

Start with a short test sentence that includes one marker. Generate the audio and listen carefully. Does the sigh sound natural? Is it in the right place? Is it too loud or too quiet? Based on what you hear, adjust the marker placement or try a different marker type. Once you find a combination that works with your chosen voice, you can explore the same pattern to longer scripts.

Keep in mind that regenerating the same text does not always produce identical audio — ElevenLabs adds slight variation to sound more natural. If a marker produces a good result, note the exact placement and voice you used so you can replicate it in future scripts. If a marker consistently produces poor results with a particular voice, switch voices or remove the marker entirely.

Frequently Asked Questions

Can I use multiple markers in one sentence?

Yes. You can write "I've been [sighs] waiting all day [breathes] for this moment" and ElevenLabs will generate both sounds. However, too many markers in quick succession can sound choppy. Test first to may support the result sounds natural rather than scripted.

Do markers work if I'm using the ElevenLabs API?

Yes, markers work the same way in the API as they do on the website. Include them in the text string you send to the API, and the generated audio will include the breathing sounds at those locations. Check the ElevenLabs API documentation for your specific integration to confirm the exact syntax.

What if a marker produces no sound at all?

Try a different voice — some voices are more responsive to markers than others. You can also try moving the marker to a different position in the sentence. If the marker still produces nothing, remove it and regenerate. Not every voice or every placement will work equally well.

Can I add other sounds besides sighs and breathing?

ElevenLabs recognizes a limited set of markers: sighs, breathes, clears throat, coughs, and laughs. These are the only built-in markers. If you need other sounds, you would need to generate the speech separately and add custom audio in an editing tool like Audacity or Adobe Audition.

How do I know if a marker is working?

Listen to the generated audio carefully. If you hear a sigh, breath, cough, or laugh at the location you placed the marker, it worked. If you hear nothing or the sound is very faint, the marker may not have worked with that particular voice. Regenerate with a different voice or marker placement to test.