How to Narrate a Character Who Whispers or Shouts

Short answer: AudioProducer.ai has no volume control. There is no whisper mode, no shout mode, and no level slider on an individual line. A whispered line and a shouted line both come out of the same places: the words you wrote, the voice you cast, the emotion tag on the line, and the silence you leave around it. This post walks through what actually works for each, and the two shortcuts that tend to backfire.

Volume is not a setting you can reach

Here is what the editor actually gives you. Per-character voice assignment, per-line speaker tagging, per-line dialogue emotion, pauses at the project, paragraph and inline level, and a sound library for music and effects. Nothing in that list is a level control. If you have been hunting the editor for a gain slider on a line, it is worth stopping the hunt now and planning around its absence.

There is a second reason volume is the wrong lever even in principle. Loudness in an exported audiobook is a property of the whole file, and retailers check it against peak and noise-floor targets before they accept an upload. Our guide to audiobook loudness and audio levels covers those numbers. A file with one line pushed well above the rest is a file that gets flagged.

What makes a line read as a whisper

A whisper registers to a listener mostly through context. The attribution carries the first half of it. "She said" tells the ear nothing, while "she said it close to his ear, barely moving her mouth" tells the ear everything before the line even arrives. Write the hush into the tag and the surrounding narration rather than hoping the voice supplies it.

The second half is space. A hush reads as a hush when the room around it is empty, so put an inline pause before the line and give the paragraph a longer custom pause than your project default. Our post on pauses and dramatic timing goes through the four levels in detail, but the practical version here is that the quiet before the line does more work than anything on the line.

Keep whispered lines short. A long hushed paragraph asks a listener to lean in for thirty seconds, and most of them will simply stop hearing it. Break it, or let the narration summarize part of it.

What makes a line read as a shout

Shouts want the opposite treatment on length and the same treatment on space. Short lines and fragments read as raised, and a beat of silence directly before the line gives it somewhere to drop into. Then let the next paragraph react. A character flinching, or a room going still: what happens after a shout is what tells the listener how loud it was.

Attach a dialogue emotion to the line itself. Emotion is set per dialogue line in the editor, so a character can be furious in one exchange and level in the next without you changing their voice. Anger and fear both push a read toward the top of a voice's range. That is inflection rather than level, and it is the closest thing to a volume control the editor has. Our guide on adding emotion to AI narration covers how far the emotion tags carry.

Skip capital letters. Capitalization is a typographic cue for a reader's eye, and leaning on it puts your shout somewhere the audio cannot see. Write the raised voice into the sentence and the attribution instead.

Cast a second voice for a character who lives at the extremes

Every voice in the library carries descriptors, and some of them are exactly the ones you want here. There are voices tagged whisper and voices tagged strong, alongside calm, expressive and the rest. You can browse and preview all of them on the Voices page on your home screen.

A character gets one voice, so this only helps if you do something slightly unusual with it. If a character spends real page time hushed, add a second character entry for those lines, give it the whisper-descriptor voice, and re-tag the hushed lines to that entry in the editor. Auto-Assign Characters will not do this for you, since it tags by speaker and your two entries are the same speaker. It is a manual pass over the lines that need it.

Preview both voices side by side on the Voices page before you commit, because the pair has to sound like one person. That is the same problem covered in what to do when two characters sound too similar, running in reverse: there you are pulling two voices apart, here you are keeping two close. If a character only whispers once in a chapter, this is more machinery than the moment deserves. Save it for the character whose hush is a running feature.

Cloning your own voice is an option for either entry, and the same rule applies as everywhere else on the Voices page: clone only voices you own or are authorized to use.

Where sound helps, and where it swallows the line

Auto-Assign Sounds analyzes the scene and places music, soundscapes and one-shot effects to match what is happening. Used around a raised voice, this genuinely helps. A shout inside a storm reads as louder because the world is loud, and the shout has to cut through it.

The failure mode runs the other way. Writers try to make a whisper feel smaller by stacking ambience underneath it, and the words disappear into the bed. The listener does not experience an intimate moment; they experience a line they had to rewind. A dense bed also lifts the overall level of the file, which is the thing the retailer specs are watching. Frame the moment with sound and keep the narration on top of it.

Audition the line inside its scene

Previewing a voice on the Voices page tells you what the voice sounds like, and almost nothing about whether your whisper landed. Volume in narration is comparative. Generate the passage with the paragraph before it and the paragraph after it, and listen to the sequence rather than the line.

Do that on a short passage rather than the full chapter. Re-generation counts against your monthly word allowance, so settle the voice, the emotion tag and the pause lengths on a page or two, then run the chapter once. The broader habits are in our guide to directing AI audiobook narration, and pacing choices around all of this are covered in controlling pacing, pauses and emphasis.

A good place to begin is the loudest scene in your manuscript and the quietest one, set up as a pair before you touch the rest of the book. The free account gives you 1,200 words a month with no credit card, which covers two scenes, and paid plans start at $39.99 a month. AudioProducer.ai hands you the finished MP3 to download; where the book goes after that is yours to arrange. If you are still assigning voices across your cast, start with giving each character a different voice.