Voiceover used to be a blunt tool. You could pick a great voice in ElevenLabs, paste in your script, and hope the delivery matched what you heard in your head. Sometimes it landed perfectly. Often it sounded flat in the one line that needed excitement, or too cheerful in a scene that should have felt quiet and serious.
That gap is exactly what emotion tags in Eleven v4 are designed to close. Instead of rewriting the same sentence ten different ways and regenerating until you get lucky, you can now leave simple acting directions directly inside your script. A tag tells the voice how to deliver the next words, much like notes a director would give an actor in a studio.
If you create videos, audiobooks, podcasts, or ads, this small feature changes your workflow. You spend less time regenerating and can build real emotional arcs instead of one flat tone. This guide explains how tags work, where to place them, and how to avoid mistakes.
Quick Answer
To add emotion tags in ElevenLabs Eleven v4, type a short bracketed cue such as [excited], [whispers], or [laughs] directly into your script text right before the words it should affect, then generate the audio as normal. The voice reads the cue as a delivery instruction, not as spoken words. Use one clear tag at a time and test short sections first.
What Emotion Tags Actually Are in Eleven v4
Emotion tags, also called audio tags, are bracketed cues placed inside the text you send to ElevenLabs. The model treats them as stage directions. The words inside the brackets are not meant to be read aloud. They shape tone, pacing, volume, and expression for the speech that follows.
Think of your script in two layers. The spoken layer is what the listener hears. The direction layer is what you add in brackets to control how that spoken layer sounds. When the two layers are written clearly, the result feels far closer to a directed performance.
Tags You Will Use Most Often
- Emotion tags: [excited], [sad], [angry], [nervous], [calm], [happy], [serious]
- Delivery tags: [whispers], [shouts], [softly], [pause], [sighs]
- Reaction tags: [laughs], [chuckles], [gasps], [clears throat]
- Pacing tags: [slowly], [quickly], [hesitates] used sparingly to change rhythm
Not every voice responds to every tag in exactly the same way. Some voices are expressive by design and take direction dramatically. Others are built for steady narration and respond more subtly. That is normal, and it is why a quick test with your chosen voice matters before you generate a long script.
What You Need Before You Start
You do not need extra software. You need an ElevenLabs account, access to a voice that works well with the Eleven v4 model, and a script you can break into short sections. If you are working on a long project, keep a copy of your untagged script as well. That backup makes it easy to compare versions and to reuse the text elsewhere.
It also helps to decide the emotional journey before you tag anything. Ask yourself where the listener should feel curious, reassured, surprised, or moved. Tagging works best when it follows a plan instead of being sprinkled randomly over every line.
Step-by-Step: How to Add Emotion Tags to Voices in ElevenLabs Eleven v4
Follow this process for your first tagged project. Once it feels natural, you will be able to tag a full script in minutes.
- Open ElevenLabs and start a new generation. Go to the text to speech area, choose your project or start a fresh one, and paste in a short test script of four to six lines rather than your full text.
- Select an Eleven v4 voice. Pick the voice you intend to use for the final piece. Preview it with plain text first so you have a baseline for how it normally sounds.
- Write your script in short, clear sentences. Short sentences are easier to direct. Long, tangled sentences give the model too many things to do at once and make tags less predictable.
- Place a tag right before the words it should affect. For example, write [excited] We finally did it! or [whispers] Do not wake the baby. The tag belongs at the start of the phrase, not at the end of the previous one.
- Use one tag per emotional beat. If a line changes feeling halfway through, split it. Write the first part with one tag, then add a second tag before the part where the mood shifts.
- Generate and listen with headphones. Pay attention to the transition into the tagged section. A good tag feels like a natural shift in delivery, not a sudden voice change.
- Adjust the wording before you add more tags. If a line still sounds flat, try stronger punctuation, a shorter sentence, or a different tag. Often [joyful] and [excited] produce noticeably different energy.
- Tag the rest of your script in sections. Work scene by scene or paragraph by paragraph. Generate each section, listen back, and only move on when the tone is right.
- Do a full final listen. Check for consistency, volume jumps, and any place where a reaction tag such as [laughs] appears without a natural reason in the text.
- Export the version you approved. Keep your tagged script saved alongside the audio so you can recreate or revise the exact performance later.
For official detail on how the platform handles expressive generation, check the guidance from ElevenLabs and its expressive voice models. It is worth revisiting when new tags and models are released, because support continues to expand.
A Simple Example You Can Copy
[calm] Welcome back to the show. [excited] We have a result I honestly did not expect. [whispers] And this is the best part. Read it without the brackets and it still makes sense — a good test for any tagged script.
How to Build an Emotional Arc, Not Just Random Effects
Beginners often tag every line. Professionals tag the turning points. Most scripts only need clear direction at the opening, at moments of surprise or tension, at emotional peaks, and at the closing line where you want the listener to remember a feeling.
Try this structure for a five-minute narration. Open calm and confident. Add curiosity before a reveal. Use a softer tag for a reflective section. Build energy for the main payoff, then return to calm for the conclusion. That rise and fall is what makes a performance feel human.
You can find more practical production ideas in our guides on NetworkUstad covering AI tools and creator workflows, where we test features in real projects rather than just listing settings.
Common Mistakes to Avoid
Stacking too many tags together. Writing [excited] [shouts] [laughs] before one short line gives the model conflicting instructions. Pick the single direction that matters most for that moment.
Tagging every sentence. Constant shifts make the audio feel unstable. Leave neutral lines untagged. A contrast between normal delivery and tagged delivery is what makes the tagged moments stand out.
Putting the tag in the wrong place. A tag placed after a sentence cannot change that sentence. Always place it before the words you want it to influence.
Using tags to fix bad writing. If a line is confusing, emotional direction will not save it. Rewrite first, tag second.
Skipping the test generation. Voices differ. A tag that sounds perfect on one voice may be too strong on another. A thirty-second test can save you regenerating a twenty-minute script.
Forgetting the listener context. A whispered line may work beautifully on headphones and disappear on a phone speaker in a noisy room. Consider where your audience will actually listen.
Tips for More Natural Results
Combine tags with punctuation. An exclamation mark already adds energy, so pair it with a milder tag. A line written as [softly] I understand. will usually sound more believable than a dramatic tag fighting against dramatic punctuation.
Also vary sentence length, and keep a personal list of tag and voice combinations that worked. That small library becomes a real time saver.
Conclusion
Emotion tags in Eleven v4 give you something text to speech has wanted for years: direction. With a few bracketed cues such as [excited], [whispers], and [laughs], placed carefully before the right words, you can shape a performance instead of accepting whatever tone the first generation gives you.
Start small. Tag a short script, listen closely, and learn how your favourite voice responds. Then build up to full projects with a clear emotional arc and only as many tags as the story needs. That simple discipline is the difference between audio that sounds generated and audio that sounds performed.
Frequently Asked Questions
Do I have to pay extra to use emotion tags in Eleven v4?
No separate tag feature needs to be bought. Access depends on your ElevenLabs plan and whether the voice and model you choose support Eleven v4. Describe pricing in general terms: free tiers are limited, and paid tiers give more characters, voices, and commercial options.
Will the voice read the words inside the brackets out loud?
It should not. Correctly written bracketed tags are treated as delivery instructions. If a word is being spoken, check that your brackets are closed properly and that the tag is a recognised, simple cue rather than a long sentence.
Can I use two emotion tags in one sentence?
Yes, if the feeling genuinely changes mid-sentence, but it is usually cleaner to split the sentence into two parts and place one tag before each part. That gives the model a clearer switch point and sounds more natural.
Why does a tag work on one voice but not on another?
Voices are trained and designed differently. Expressive voices respond strongly, while steady narration voices may respond subtly by design. Always test your chosen voice with a short tagged sample before a long generation.
What are the best tags to try first?
Start with [excited], [calm], [whispers], [sad], and [laughs]. Those five cover most everyday needs, are easy to judge by ear, and teach you quickly how strongly your voice takes direction.
Leave a Reply