AI Radio Play Generator

What a radio play needs from a generator

A radio play has no picture. Everything the audience knows about the place, the people and the danger comes through the ear: who is speaking, how they sound, what they say about their surroundings. That puts three demands on a tool: distinct voices for every character, a narrator who can set the scene, and control over how lines are delivered.

AudioDrama covers those three. You write a plain-text script, each character gets one of 16 studio voices, a Narrator reads any line without a name, and square-bracket directions shape the delivery. It does not mix sound effects or music; that stays with you and your editor.

Radio play conventions that work with AI voices

These are practical habits from writing for audio, adapted to a generated cast:

• Say where we are. The listener cannot see the ferry, the fog or the lamp. Have the Narrator or a character mention it early: “Fog's thick as wool tonight.”
• Use names often. Characters addressing each other by name (“Keep the lamp lit, Tom”) tells the listener who is who without a face.
• Keep the speaking cast small. Three to five distinct voices per scene is plenty. More than that, and listeners lose track unless the voices are very different.
• Contrast the voices. Cast a gruff older voice against a young nervous one. The built-in voices include narrators, young and older voices, and British, Irish and Australian accents.
• Write silence on purpose. A [short pause] inside a line makes room for a reveal. Silence cannot be a line of its own, so attach it to the line it belongs to.
• Put the hook at the end of the scene. A bell that rings from nowhere is a cliffhanger. End scenes on a question or a sound.
• Describe sounds with words. Since the tool cannot make a knock, make a character hear one: “Something is knocking on the hull.” You can then add the real knocking in your editor.

An example: the opening of a radio play

This is the opening of an imagined play, written in the format AudioDrama reads. It has four voices: a Narrator and three characters. It is 534 bytes once speaker tags are counted, so at least 36 seconds of credit.

Narrator: Scene one. The night ferry, halfway across the bay.
Captain Hale [gruff, tired]: Fog's thick as wool tonight. Keep the lamp lit, Tom.
Tom [young, nervous]: Captain... do you hear that? Something is knocking on the hull.
Captain Hale [low, suspicious]: [short pause] Nothing out there but water, lad.
Elsie [calm, older]: Forty years I've crossed this bay. Water doesn't knock.
Tom [whispering]: Then what is it?
Narrator: The knocking stopped. Then, from somewhere above them, a bell began to ring.

A few things to notice:

• The Narrator opens with Scene one and the location, then returns at the end for the hook.
• The speaker Captain Hale has a two-word name. Names can be up to five words, as long as they contain no brackets or colons.
• Each direction is brief and tells the voice something different: tired, nervous, suspicious, calm, whispering.
• There are no FX: or MUSIC: lines. They would be read aloud as if they were characters, so sound cues belong in your notes, not in the script you paste. The audio drama script format guide shows how to convert them.

From scene to episode

Because one job is capped, think in scenes. A workable routine:

1. Write each scene as its own script. Give it a title in the app, such as “Ep1 Sc1 ferry”.
2. Keep character spellings identical across scenes so the same voice can be chosen each time. Names are case-sensitive.
3. Generate each scene, listen, and edit directions where a line misses. Generation is a paid step once your free minutes are used, so test the tricky lines first.
4. Download the MP3s from your library and join them in an audio editor. This is also where you add ambience, effects and music.

The sibling guide on how to make an audio drama walks through the whole process, from premise to publishing.

What it does not do

• No sound effects, no music, no mixing.
• No overlapping dialogue: lines play one after another.
• No voice cloning and no uploaded voices: 16 curated English voices only.
• No PDF, FDX or Fountain import. Paste plain text.
• Up to 16 characters per script (the Narrator counts as one).
• 2 free minutes on sign-up; then $9 for 60 minutes or $29 for 240 minutes, one-time. See pricing.

If you want to write dialogue that really performs, the page on text to speech with emotions explains how directions work.