Can I turn a sound effect or music bed down in AudioProducer.ai?

No. There is no per-cue volume slider, no fader, and no ducking. You control a scene through placement (where a cue sits and how much text it covers) and selection (which sound you pick). If a passage is too loud or too busy, the fix is to remove a cue, shorten the stretch it runs under, or swap it for a sparser sound.

How do I keep background music from covering the narration?

Choose a bed without a melody competing in the same register as the voice, run it under fewer paragraphs than you first want, and stop it before the dialogue that matters most. Because there is no ducking, the separation has to come from what you picked and where you put it, not from riding a level.

Should I use Auto-Assign Sounds or place every cue by hand?

Use Auto-Assign Sounds as a first pass, then subtract. It reads the scene and places music, ambient soundscapes, and one-shot effects from the library, which is faster than hand-placing everything. It is deliberately generous, so the human pass that follows should be about cutting what decorates and keeping what carries information.

How many sound cues should a chapter have?

Fewer than most first drafts contain. Nearly every overloaded scene has too many cues rather than cues that are too loud. Build a cue sheet from the manuscript before you open the editor so you can see the density while it is still cheap to change, then run one subtraction pass on the finished chapter before you export.

Sound Design for Audio Fiction

Sound is the part of audio fiction that people notice last and judge first. A listener will not tell you that the rain under chapter four sat too close to the dialogue; they will tell you the chapter was hard to follow. Most sound-design questions we get are really one question wearing different clothes: how do I get the backdrop to sit behind the voices? This page is the orientation for that whole subject, and it starts by being honest about the shape of the answer in AudioProducer.ai, because the shape is not what most people expect.

What sound is doing in a scene, and what it is not

A soundscape is not decoration you add once the writing is finished. It carries information the prose no longer has to carry. When a scene opens on a market and the listener hears the market, the narrator does not have to describe the market, and the first line of dialogue can be the first line of dialogue. That is the trade sound is making on your behalf: it buys back the sentences you were spending on establishing shots.

It also marks edges. Print has white space, chapter breaks, and a page you can see the end of. Audio has none of that. A change in the ambient bed is one of the few ways a listener can tell that the ground moved without being told. Silence is the other, and it is the more underused of the two.

What sound is not doing is fixing a scene that does not work. If two characters read as interchangeable, thunder will not separate them. If a paragraph is confusing on the page, a storm behind it makes it confusing and loud. Sound amplifies whatever structure is already there, including the absence of one.

The two levers: placement and selection

Here is the part worth reading twice, because it determines how every other question on this subject gets answered.

AudioProducer generates audio from cues attached to moments in your text, and it plays audio from the sound library, both the built-in library and whatever you have uploaded into your own. You choose which sound goes at which moment. You do not get a mixing desk. There is no per-cue volume slider, no fader, and no ducking that automatically pulls the backdrop down when a character speaks.

So the levers are two, and neither of them is a level:

Placement - where in the text a cue sits, how long the stretch of text under an ambient bed runs, and what falls in the gaps between cues.
Selection - which sound you pick. A distant, sparse, low-detail sound sits behind dialogue in a way a close, busy, high-detail one never will, no matter where you put it.

This is not a limitation to work around so much as a different set of controls to work with. A film mixer solves the crowded-scene problem by riding a fader. You solve it by choosing a quieter room and putting fewer things in it. The result can be just as clean; the decisions simply happen earlier, in the choosing rather than in the mixing.

Removal is the volume control

The practical consequence: when a passage is too loud or too busy, the fix is to take something out. Not to turn it down, because there is nothing to turn. Cut the cue, shorten the stretch it covers, or swap it for something with less going on.

This sounds like a downgrade until you use it for a while. Nearly every overloaded scene we see has too many cues in it, not cues that are too loud. A market with footsteps, chatter, a cart, a bell, and a dog is not five times as vivid as a market with chatter and a bell; it is a wall. The subtraction pass, done once before you export, is the single highest-return habit in this whole subject.

The same logic covers the question people ask most often, which is how to keep music from swallowing narration. Pick a bed with no melody competing for the same register as the voice, run it under fewer paragraphs than you think you want, and let it stop before the dialogue that matters most.

Plan the cues before you open the editor

Sound decisions made while you are looking at the editor tend to be reactive: you see a paragraph, you think of a sound, you place it. Do that through a chapter and you get a cue list with no shape, dense in the places you happened to be paying attention.

The alternative takes about twenty minutes per chapter. Read the manuscript once and write down the sounds the text already names out loud. Read it a second time for the sounds the scene implies but never says, and for the silences worth protecting. That list is a cue sheet, and it is the artifact that turns sound design from a thousand small choices into a handful of deliberate ones. It also tells you your density before you have committed to anything, which is when density is cheap to change.

Auto-Assign Sounds, and what to do with what it gives you

Auto-Assign Sounds reads the scene and places music, ambient soundscapes, and one-shot effects from the library to match what is happening. Storms get thunder and wind. It is the fast version of that first read-through, and it is a genuinely good starting point on a first chapter.

Treat it as a draft, not a delivery. The pass it produces is generous by design, and generous is the wrong setting for a scene where two people are talking. Run it, then run subtraction over the result: keep what carries information, cut what is decorating. The combination of an automatic pass and a human subtraction pass is faster than hand-placing everything and ends up sparser, which is the direction you want to be wrong in.

Where it all gets checked: the render

Generate the chapter and listen to it end to end without the text in front of you. Reading along hides problems, because you already know what the line is going to be. What you are listening for is narrow: any moment where you lose a word, any bed that outstays the scene, any effect that arrives so precisely on its noun that it reads as a sound-effect joke.

Everything you find has the same two fixes available. Move it, or swap it, or take it out. That is the whole toolkit, and once you have run a chapter through it a couple of times, it stops feeling small.

The sound-design series, grouped by what you are working on

Every piece below goes deeper on one part of the above. Start with the group that matches what is in front of you.

Start here, by format. The two end-to-end walkthroughs, plus the two habits that make everything after them easier.

How to add sound design to an audio drama - the full pass for multi-voice fiction, from spotting a scene to exporting the finished file.
How to add sound effects and music to an audiobook - the same job in a single-narrator book, where restraint matters more and the listener expectations are different.
How to add ambient sound to an audio story - the distinction between a bed that runs under a scene and a one-shot effect that marks a moment, which is the distinction everything else rests on.
How to build a sound effects cue sheet from your manuscript - the two-pass read that gets your cue list out of the editor and onto paper before you commit.

Specific scenes. The four that come up most, each with a worked pass.

How to sound design a storm or weather scene - why rain is a bed and thunder is not, and how weather behaves when it has to sit under dialogue.
How to sound design a battle or action scene - choosing three cues instead of thirty, and putting the silence before the loud moment rather than after it.
How to create crowd noise in an audio scene - building a room the ear reads as full, when lines render one after another rather than overlapping.
How to use sound for scene transitions in an audio drama - marking a scene break for a listener who cannot see the white space.

Music. Beds under prose, and the rights question that comes with them.

How to add background music to an audiobook - where a bed earns its place, and how to keep it from competing with the narration.
Music licensing for an AI audiobook or podcast - what a license has to cover before you publish, and why an old song is not an old recording.

Restraint and delivery. Knowing when to stop, and what the file has to meet once you do.

When not to use sound effects in an audiobook - the subtraction pass in full, and why a named sound is not automatically a cue.
Audiobook loudness and audio levels - the RMS, peak, and noise-floor targets retailers check, and what to do with a rejected file.

If you are building the drama itself rather than its soundscape, the wider walkthrough is how to make an audio drama with AI. Everything outside this series sits in the full guide index.