Eleven v4 如何为 AI 配音添加停顿

Blog; Resources

ListenListen to this article

On this page

With an AI Text to Speech model, delivery is everything. While the quality of a voice and an understanding of prosody go a long way, without timing, a sentence falls apart. An extra beat after a line or an extended silence can be instrumental in creating a high-quality output.

Generic Text to Speech tools leave pauses up to chance. One generation might capture what you’re looking for, only for it to be gone the next time you hit regenerate. Eleven v4 offers improved regeneration workflows, making that a thing of the past. Fine-tune individual lines or phrases with ease in your script.

And, to build the very best script possible right from the start, Eleven v4 gives you three ways to build pauses in AI voiceover directly into your text. Here’s what each one does and when you should use them.

Hear Eleven v4 in action

Summary

SSML tags don’t work in Eleven v4. The system uses Audio Tags, where you write stage directions like [long pause] or [whispers] directly into your script; Altering punctuation can play with the rhythm of a sentence, with dashes, ellipses, and short sentences shaping how your TTS output sounds; To write effective pauses, place them where a human speaker would naturally stop; When writing long-form audio, use the same pause conventions throughout to help keep your output consistent

What changed from earlier models?

Earlier ElevenLabs models, like Multilingual v2, supported SSML syntax. You could write in a pause during a sentence with a structure like to add in a 2-second pause. Eleven v4 continues the approach introduced in Eleven v3 by replacing SSML tags with a much more intuitive alternative, Audio Tags.

For writers, the replacement makes creating pauses in your AI voiceover much easier. Instead of having to know what exact SSML tag to use, you just need to write in plain language, with [pause] and [long pause] doing the work for you. These tags become directions for the model, reacting to them and using the context of the scene to create a nuanced final performance. The exact length of the pause fits the moment, rather than being a pre-defined, mechanical silence gap.

Though it's not the focus of this article, you can also use Audio Tags to create emotional Text to Speech output by varying emotions across a paragraph.

The three pause controls in Eleven v4

Eleven v4 gives you three ways to add pauses in AI voiceover.

We’ll show how each of these works in practice in our text editor and outline how they feel with an example audio.

[pause]

A pause is a short beat. It’s roughly equivalent to a breath, or the natural space a speaker takes before continuing with a sentence. The best place to use the [pause] Audio Tag is after a setup line, between two ideas you want to separate, after a phrase that needs extra emphasis, or anywhere a period alone feels too rushed.

In the example below, we use the same line with and without the [pause] Audio Tag. Even without hearing the audio clips, you'll notice that one lasts four seconds and the other lasts six. The extra beats that [pause] delivers add up in the final clips.

With [pause] example

No pauses example

The clips above were generated with Flint on Eleven v4.

[long pause]

A long pause is more deliberate and will create moments of silence in your audio clips. You can use that silence for dramatic tension, during a break between scenes, before a reveal, or anywhere you want to emphasize silence.

Here's an example of a [long pause] in a scene.

This audio clip was voiced by Spuds Oxley on Eleven v4.

Punctuation

Eleven v4 uses punctuation as a pacing instruction. Adding an ellipsis will slow down the delivery of a line, allowing you to create moments of hesitation. Changing up your sentence structure will break up the monotony of a longer passage. Short. Snappy. Direct.

While you can use Audio Tags throughout your script (and should to improve the delivery of your performance), punctuation alone can go a long way.

In the following passage, you can see just how punctuation brings a text to life, even without Audio Tags.

This audio sample was brought to life with Lauren B in Eleven v4.

Pauses in long-form audio

Generating long-form audio works best when you take a consistent approach across the entire length of your script. For example, if you only use punctuation for the first 1,500 characters and then start introducing [long pause] Audio Tags in the final 1,000 characters, the output might sound strange.

Before writing your script, decide what approach you want to take. Or, if you’ve already written it, be sure to apply Audio Tags consistently throughout. The ‘v4 Enhance’ button next to ‘Generate speech’ will automatically add Audio Tags to your writing.

Eleven v4 brings your scenes to life.

By controlling the pacing of your text and adding Audio Tags, you can create rich vocal performances with over 17,500 voices from our Voice Library.

Adding pauses in AI voiceover FAQ

Can you use SSML break tags in ElevenLabs

Eleven v4 doesn’t support SSML tags. Natural language descriptors like [pause] and [long pause] give you a more intuitive solution. You add pauses to your text the same way you layer emotional direction by using Audio Tags.

How long is a [pause] in Eleven v4?

A pause adapts to the surrounding context and the voice you’ve chosen. Moving away from a fixed-length pause system allows for more expressive and natural pauses in your audio outputs.

Can I add pauses in ElevenAPI audio outputs?

You can add pauses to scripts you use to generate audio via ElevenAPI by using any of the methods outlined in this article. Use Audio Tags, punctuation, or allow Eleven v4 to contextually add pauses based on the context of your writing.

Jack Limebear is on the Growth team, acting as a content writer and strategist across the blog and insights pages. Before ElevenLabs, he spent over a decade leading content strategy for organizations ranging from fast-growing SaaS startups to Fortune 500 companies. He holds a Master's degree in English Literature from the University of Cambridge.