How to make an emotional AI audiobook: character voices and emotion tags

Short answer: cast a voice for each main character, then open Customize voice and describe the character in one plain sentence, like "a gruff old fisherman, tired and kind, speaking low". Set the pace and preview it. Then, for the moments that need it, type [ in a sentence to add an emotion tag such as [sigh], [laugh] or [whispers]. The voice gives each character their personality; the tags give a scene its emotion.

What makes an audiobook sound expressive

Listeners rarely complain that a narrator wasn't emotional enough on a single line. What they notice is that everyone sounds the same: the villain, the child and the narrator all speak in one voice at one pace. That flattens a story more than anything else.

Good human narrators fix this with characterization rather than with big emotion on every sentence. Each character gets a recognizable age, manner and speed, and the narrator stays steady underneath them. The feelings come from the writing, carried by a voice that suits the person speaking.

Then they save the big emotional reactions for the moments that earn them: the sigh before bad news, the laugh that breaks the tension, the line that's whispered.

Warblize gives you both tools. Customized voices set who each character is, once, for the whole book. Emotion tags mark the individual moments in your manuscript.

Step 1: Choose your narrator

Your book starts with one voice, the narrator, shown as Default in the Characters panel of the studio. Every sentence you don't assign to a character is read by this voice.

Pick a narrator that suits your genre and leave it close to natural. A calm, clear narrator makes the character voices stand out. If you're unsure, browse the voice library and listen to a few previews first.

Step 2: Add your characters

Open Characters in the studio and press +. Choose a voice from the library, then use the ⋯ menu to Rename it to the character's name, such as "Old Tom" or "Detective Ruiz".

To skip this step and the next, press Find characters instead: it casts the main characters and assigns their lines for you. See Find characters.

Cast only the characters who speak often. Three to five distinct voices is plenty for most novels. A minor character with two lines can usually stay with the narrator.

When choosing a voice, start from the closest match in gender, rough age and language. The description shapes how a voice performs. It doesn't turn a young voice into an old one or a female voice into a male one, so a good starting voice matters.

Step 3: Give each character their lines

In the manuscript, click the voice avatar beside a sentence and choose the character who speaks it. Do this for the character's dialogue.

Many authors leave dialogue tags ("she said quietly") with the narrator and give only the spoken words to the character. It sounds like a narrator performing a scene, which is what listeners expect.

Step 4: Customize the voice

Hover over a character and click the settings icon. Two controls open under the character.

Pace runs from Very slow to Very fast, with Natural in the middle. Small changes go a long way:

Keep Very slow and Very fast for characters who are defined by it. Across a whole book, extreme pace gets tiring to listen to.

Describe the voice takes up to 300 characters of plain text. Write it the way you'd brief a narrator: who the character is, how they sound and how they speak.

How to write a voice description that works

The best descriptions are one sentence with three parts: who, how they sound, and how they speak.

• A gruff old fisherman, tired and kind, speaking low
• A bright, curious eight-year-old girl, quick and excited
• A cold, calculating villain who speaks softly and precisely
• A warm grandmother telling a bedtime story, gentle and slow
• A confident young detective, dry and a little sarcastic

A few rules make a big difference.

Describe the person, not the plot. "A widow who has stopped expecting good news, quiet and flat" works. "She is sad because her husband died in chapter two" gives the voice nothing to perform.

Give the character's usual emotional state. The description applies to every line the character speaks, so choose the feeling they carry through most of the book, like weary, anxious, cheerful or guarded. Leave the moment-to-moment changes to your writing.

Don't contradict yourself. "Shy but loud and commanding" splits the difference and sounds like neither. Pick one.

Keep the narrator light. If you customize the narrator, use something gentle, such as "a calm, warm storyteller, unhurried". The characters should carry the color.

Stay within the voice's language. An English voice can take "a soft Irish lilt" or "a clipped, formal manner". For a character who speaks another language, start from a voice in that language.

Step 5: Preview, adjust, and preview again

Press Preview to hear the voice read a short line with your settings. Your changes are saved as you type, so there is no Save button. If it isn't right, change one thing at a time: a word in the description, or one step of pace. Then preview again.

If a description overshoots (too gravelly, too whispery), soften it with words like "slightly", "a little" or "a hint of". Reset clears both controls and returns the voice to how it sounds in the library.

Settings are saved for that character in that book only.

Step 6: Add emotion tags where a moment needs it

Click a sentence in the manuscript to edit it, put the cursor where the emotion should happen, and type [. A picker opens with more than 30 emotions and sounds:

Pick one, or type a few letters to filter the list. The tag appears in the sentence as a colored chip, like [sigh], and the voice performs it at that point instead of reading it aloud. If what you need isn't listed, type it yourself and choose it under Your own, for example [nervous laugh] or [sharp intake of breath].

Where to put a tag

Put it right before the words it affects. "[sigh] Nobody came." sighs, then speaks. "Nobody came. [sigh]" speaks, then sighs. Both are useful; they just mean different things.

Use [whispers] at the start of a whispered line, and keep the whole line short. Long whispered passages are hard to listen to.

Use pauses for timing. A [short pause] before a reveal or the punchline of a joke does more than any emotion tag. A [long pause] works at the end of a scene.

Keep them for the moments that matter. A chapter with a tag on every line sounds like overacting, which is the opposite of what you want. A few in each chapter, at the emotional turns, is usually right. Let the voice description carry the character's mood the rest of the time.

Put tags on the right voice. A [sob] on a narration line is performed by your narrator. If it's the character crying, make sure that sentence is assigned to the character.

Adding or removing a tag changes the sentence, so it needs new audio. Press play on it to generate the new version and hear how it sounds.

Step 7: Generate

Generate your chapters as usual. Every line assigned to a customized character is read with its description and pace, every tag is performed where you placed it, and the narrator reads the rest.

Customize before you generate. Settings apply to audio generated after you change them. Lines that already have audio keep it, so set up your cast first, then generate the chapters they appear in.

Then listen through one chapter with several characters in it before you generate the whole book. It's the cheapest point to catch a voice that doesn't fit.

A quick checklist

1. Pick a calm narrator that suits your genre.
2. Cast three to five main characters with the closest matching voices.
3. Assign each character's dialogue in the manuscript.
4. Describe each character in one sentence: who, how they sound, how they speak.
5. Set pace sparingly.
6. Add emotion tags at the emotional turns of each scene, not on every line.
7. Preview, adjust, then generate one chapter and listen before doing the rest.

A well-prepared manuscript helps as much as the voices do. See how to prepare a manuscript for an audiobook for the formatting that reads cleanly aloud.