How to Write Dialogue for an Audio Drama

To write dialogue for an audio drama, write it for a listener who cannot see anything. Name characters early in each scene, let reactions and sound carry physical action instead of having characters narrate what they are doing, and give every character their own rhythm and vocabulary so the lines stay distinguishable before a single voice is cast. Then read it aloud, or produce a rough draft and listen, and cut every line that only makes sense on the page.

This guide is about the writing. If you want the page layout (cues, SFX lines, how a script is set out), see how to format an audio drama script. If you are narrating an existing novel with lots of conversation rather than writing a drama, how to narrate dialogue-heavy scenes is the better fit.

Why audio dialogue does the work a camera would do

On screen, dialogue is one layer among many. The camera shows who is in the room, where they stand, what they are holding and how they look at each other. The words can be sparse because the picture carries the rest.

In audio, the words, the voices and the sound are all there is. Every piece of information the listener needs has to arrive through one of those three channels, and dialogue is the one the writer controls most directly. That means audio dialogue is doing three jobs at once: it tells the story, it tells the listener who is speaking, and it tells them what is happening around the speaker. Good audio writing does the second and third jobs quietly, so the listener only notices the first.

The failures are predictable. Either the writer forgets the listener is blind, and the scene turns into voices in a void, or the writer remembers too hard, and every character starts describing the room. The rest of this guide is about staying between those two.

Names in the first exchanges, and how not to overdo it

A listener cannot glance at a character name above a line. They learn who is who by connecting a voice to a name, and they need to hear the name to do that. So in the first few exchanges of any scene, at least one character should address another by name, the way people naturally do when they greet someone, interrupt them or get their attention.

A workable rule: name each speaker within their first two or three lines in a scene, and again whenever someone has been silent for a while and comes back in. After that, stop. Once a voice is attached to a name, repeating the name in every line sounds like a hostage negotiation.

Some habits that keep naming natural:

• Use names where people really use them. Greetings, arguments, warnings and calling across a room all invite a name. Calm two-person conversation mostly does not.
• Let a third character do the introducing. "Maya, this is the man I told you about" names two people at once, and it sounds like a line rather than a label.
• Use relationships and titles. "Mum", "Sergeant", "Doctor" and "boss" tell the listener who is speaking and what the relationship is, which is double the information for one word.
• Keep scene casts small. Every extra speaker in a scene is another name the listener has to learn. We cover how many speaking parts a scene can hold in how many voices does an audio drama need, and why a first episode should be stricter still in how to write an audio drama pilot episode.

Signposting action without "As you can see, I am opening the door"

This is the classic audio-drama tell. The writer knows the listener cannot see the gun, so a character says "Put down that gun you are holding in your right hand." Nobody talks like that, and listeners hear it immediately.

Three cleaner fixes, roughly in order of how often you will use them:

• Let another character react. The person who sees the gun says what a frightened person would say: "Whoa. Okay. Nobody needs to get hurt." The listener now knows there is a weapon, who is holding it and how scared the room is, and no one has described anything.
• Let the speaker's intention reveal it. Instead of "I am opening the door," the character says "Stay behind me" and then "It's empty." The action is implied by what they want, not narrated.
• Let a sound do it. A latch, a creak and a change in room tone say "the door is open" faster than any line. Dialogue then only has to handle what the sound cannot, such as what is behind the door.

The third fix is where the script meets production. Turning stage directions into sound cues, step by step, is covered in how to turn a screenplay into an audio drama, and planning those cues across a whole script in how to build a sound effects cue sheet. For the writing, the useful test is simple: if a line exists only to describe something visible, try giving that job to a reaction or a sound, and delete the line if either one works.

Distinct speech patterns as casting help

Voice casting separates characters by sound. Writing can separate them before casting even starts, and it does the heavier lifting. If two characters speak in the same rhythm with the same vocabulary, no amount of voice contrast fully fixes it, because listeners track meaning as much as timbre.

Give each main character a pattern you could recognize from a transcript with the names removed:

• Rhythm. One character speaks in short, clipped lines. Another runs sentences together and trails off. Put them in the same scene and the listener can tell them apart with their eyes closed, which is the whole point.
• Vocabulary. A doctor, a teenager and a retired sailor reach for different words for the same thing. Pick a handful of words each character uses and a handful they never would.
• Habits. One character answers questions with questions. One always has to have the last word. One never says anyone's name. Habits like these become part of how the listener recognizes the speaker.

Accent and dialect are a separate tool with their own risks. If you are thinking about written-out dialect, read how to handle dialects in an AI audiobook first. When two characters still blur after casting, what to do when two characters sound too similar covers the fixes on both the casting and the text side.

Read it aloud, then listen to a draft

Dialogue that reads well can still sound wrong. Long sentences with nested clauses lose the listener halfway through. Two characters whose names start with the same sound get confused. A joke that depends on spelling does not land. The only reliable check is to hear the scene.

Reading aloud catches most of it. Hearing it in separate voices catches the rest, especially the question of whether the listener can follow who is talking. With AudioProducer.ai you can paste a scene as text or upload an EPUB, let Auto-Assign Characters make a first pass at who says which line, then correct the tagging and pick a voice for each character from the Voices page, where you can preview them. Auto-Assign Sounds can suggest effects and ambience for the scene, which is a quick way to test whether a sound really carries the action you took out of the dialogue. Pauses let you control the gaps between lines. Then download the MP3 and listen somewhere you are not looking at the script.

Listen for three things: any moment you were unsure who was speaking, any line where a character described something a reaction or sound could have shown, and any stretch where two characters sounded like the same person. Each points back to one of the sections above. When you are happy with the audio, what you do with it is up to you: AudioProducer makes the audio file and does not publish or distribute anything for you.

A quick dialogue checklist before you produce

• Every speaker is named within their first few lines of each scene.
• No line describes something only a camera could show, unless the plot needs it said.
• Physical action is carried by reactions, intention or sound cues.
• Each main character has a rhythm and vocabulary you could recognize without names.
• No scene has more speakers than a listener can hold.
• You have heard the scene, not only read it.

For more on writing and producing audio drama, browse all of our articles, including how to make a radio drama with AI voices.

The fastest way to find out whether your dialogue works is to hear it. Take one scene you have already written, cast two or three voices, and hear your scene read in separate voices on the free tier, which gives you 1,200 words a month to try it.