How to Add Emotion to Text to Speech

Most AI text to speech reads a joyful line and a heartbreaking one exactly the same way. That flat, one-note quality is the fastest way to tell a listener they are hearing a machine, and it is the single biggest reason narration, voiceovers, and character dialogue fall flat.

With Voice Creator Pro, when you use the Qwen3 TTS model, you can now generate AI speech with emotional nuance, like, sarcasm, anger, fear, and more. You can assign an emotion to your text from a dropdown of 13 emotions, each with five intensity levels, and the voice you picked performs the line with that feeling. Crucially, this is not a property of one specific voice. You choose the voice first, whether that is a voice from the library, a voice you cloned from a short recording of your own, or an audio sample you uploaded, and then you tell it how to feel. The same voice can sound calm in one line and furious in the next.

This guide shows you how to do it, with audio you can play right now.

Hear the difference

The clips below are all the same voice, "Aria," which was cloned from a short reference recording. The first is the voice at rest. The rest are that same cloned voice with an emotion assigned to the text. Nothing about the voice identity changed between them. Only the emotion did. Press play to hear it.

That range, from one cloned voice, is what emotion control gives you.

What you can control

There are two dials: the emotion itself and how strongly it comes through.

The 13 emotions

Qwen3 TTS gives you 13 emotions to assign, on top of a neutral baseline. Each one suits different work:

Emotion	Good for
Happy	Upbeat ads, cheerful explainers, friendly assistants
Angry	Villains, confrontations, high-stakes drama
Sad	Emotional narration, reflective scenes, tragedy
Fearful	Horror, thrillers, tense game moments
Whisper	Secrets, intimacy, suspense, ASMR-style reads
Excited	Trailers, reveals, hype moments, reactions
Tender	Comfort, romance, bedtime and children's content
Dramatic	Cinematic narration, epic storytelling, big beats
Calm	Meditation, wellness, guidance, onboarding
Authoritative	Announcements, training, commands, documentaries
Sarcastic	Comedy, character voices, dry commentary
Playful	Kids' content, games, lighthearted social videos
Storytelling	Audiobooks, fables, long-form narration

Five intensity levels

Every emotion has an adjustable intensity so you can go from a subtle hint to a full-force delivery: very mild, mild, normal, strong, and extreme. A "sad" at very mild is a faint touch of melancholy. A "sad" at extreme is a voice cracking between words. The right level is usually the one a real person would actually use in that moment, not the loudest one available (more on that below).

How to add emotion to your text, step by step

1. Select the Qwen3 TTS model. Emotion controls are only availabel for the Qwen3 TTS model right now, so the emotion dropdown only appears when Qwen3 TTS is the selected model when generating speech in the voice clone tab.
2. Choose your voice. Pick any voice you want the line spoken in: a built-in library voice, a voice you cloned from a few seconds of your own recording, or an audio sample you uploaded. The emotion applies to whichever voice you select.
3. Enter your text. Type or paste the line you want spoken.
4. Pick an emotion from the dropdown. Choose one of the 13 emotions.
5. Set the intensity. Slide from very mild to extreme depending on how strong you want the feeling.
6. Generate. The voice reads your line with that emotion. Not happy with the read? Change the emotion or nudge the intensity and generate again.

Because you select the voice separately from the emotion, you can keep one consistent narrator across an entire project and simply change how it feels line by line.

Emotion in the Audiobooks tab

For long-form work, you do not want to generate line by line. Voice Creator Pro's Audiobooks tab lets you assign emotion across a whole document. With Qwen3 TTS selected, you have two options:

• Highlight and assign. Select a word, phrase, or passage inside a segment and assign an emotion to just that selection, so a single line can turn tender or dramatic exactly where you want it.
• Assign per segment. Set an emotion for an entire segment from the dropdown, so a whole paragraph or scene carries one mood.

This is what makes emotion control genuinely useful for a book rather than a demo. A chapter can open calm, build to something dramatic at the climax, and settle into tenderness at the end, all in the same narrator's voice. For the full workflow, see the guide on how to generate an audiobook.

Getting a believable emotional read

A few habits separate a voice that sounds acted from one that sounds real.

1. Start from a neutral baseline. Generate a plain read first so you know how the voice sounds at rest, then add emotion on top. It is far easier to judge how much feeling to add when you can hear it against the baseline.

2. Match the text to the emotion. Qwen3 TTS also reads the words themselves, not just the emotion setting. An "angry" tag on a polite sentence underdelivers, so give an angry line something worth being angry about. When the words and the emotion agree, the read is far more convincing.

3. Do not max the intensity. A maxed-out emotion reads as fake. Aim for the level a real person would use in that moment. A restrained, believable delivery beats a shouted one for almost every use case except deliberate extremes.

4. Or clone a reference that already carries the feeling. Emotion control and cloning are separate tools that combine well. If you clone a 3 to 10 second reference clip that already sounds passionate or tearful, that feeling carries into the cloned voice before you add any emotion setting at all.

Where emotional voices matter

• Audiobooks and drama. Narration that shifts tone between characters and scenes keeps listeners engaged across hours of content where a flat voice loses them.
• Voiceovers and video. Voiceover that matches the mood of a scene, whether a tense thriller beat or a heartfelt moment, sells the tone as much as the visuals do.
• Games and characters. NPCs and heroes get real range, from a furious boss taunt to a frightened companion, which is what makes a character feel alive rather than read from a script.
• Social and shorts. Reactions that actually sound excited, shocked, or heartbroken give short-form content energy that flat narration cannot.
• E-learning and explainers. A voice that sounds genuinely enthusiastic about the material makes training and course content far easier to sit through.

Try it yourself

Emotion control is available on the Qwen3 TTS model in both Voice Creator Pro Cloud, which is free to start with no card required, and the desktop app. Pick a voice, assign an emotion, and you can hear it in a couple of minutes.

For a fuller tour of every way to make a voice expressive, including ready-made expressive voices and voice design, see the emotional text to speech page. If your voice sounds flat for other reasons, the guide on why TTS sounds robotic and how to fix it covers the rest.

Try it free in your browser.