AI Dialogue Generator: Multi-Speaker Scenes in One Pass Fish Audio
Two or more characters in one piece of audio: how to write a dialogue script, direct each speaker, keep voices consistent across episodes, and plan credits. Start from $9.
Single-voice narration has one problem: keeping one voice interesting. Dialogue has the opposite problem: keeping several voices distinct, consistent and correctly attributed. A listener has to know who is speaking within the first syllable, without a caption.
This guide covers the dialogue script format, the casting sheet, and what multi-speaker scenes cost.
One pass versus separate clips
There are two ways to produce multi-speaker audio:
- One pass — the whole scene is generated as a single piece of audio, with speaker labels in the script. This keeps the timing and interruptions natural, and it is the right choice for scenes where characters talk over or react to each other.
- Separate clips — each character is generated alone and the clips are edited together. This gives frame-level control and is the right choice when the audio has to sync to picture, or when each line needs its own file for a game engine.
Most dialogue projects use one pass for the scene and separate clips for pickups. Decide before writing: the script format follows the production method.
How to write a dialogue script
- Label every line.
MARA:andDEV:at the start of each line, consistently. Do not alternate spellings. - Break between speakers. One line per speaker, a blank line between turns. Paragraph-shaped dialogue reads as narration.
- Put directions in square brackets.
[whisper],[laughs],[angry]at the start of the line where the emotion matters. - Keep directions out of the spoken text. Anything not in brackets is read aloud; anything in brackets is a direction.
- Write interruptions as interruptions. A dash at the end of a line, then the next speaker, reads more naturally than a full stop and a polite pause.
The casting sheet
The casting sheet is the document that keeps a series consistent:
| Character | Voice | Pace | Notes |
|---|---|---|---|
| Mara | mid-range, warm | measured | the anchor of the scene |
| Dev | lower, quicker | fast | interrupts, reacts |
| The narrator | neutral | slow | only at the top and tail |
Assign the voices once, save them, and reuse the sheet for every episode. When a new character arrives, add a row before writing the scene.
Where dialogue audio gets used
- Audio drama and fiction podcasts — the closest cousin of the podcast workflow, with a full cast instead of two hosts.
- Training and compliance scenarios — a customer and an agent, a manager and an employee, voiced as a real conversation.
- Game chatter and barks — pairs of NPCs talking to each other; the game voice guide covers the engine-side workflow.
- Language-learning dialogues — two speakers at a natural pace, with clear turn-taking.
- Character demos — a short scene that shows what a cast sounds like before a full production; the character voice guide covers casting.
What does a dialogue scene cost?
One credit covers about 1,000 characters of the scene, not per speaker:
| Scene | Characters | Approximate cost |
|---|---|---|
| 2-minute two-person scene | ~1,800 | 2 credits |
| 5-minute scene | ~4,500 | 5 credits |
| 10-episode series, 5 minutes each | ~45,000 | ~45 credits |
A series fits comfortably in the $29 Creator pack, which also covers monetized release. If the dialogue is generated by a game or an app at runtime, the API guide explains keys and credits.
Frequently asked questions
Can the engine generate several speakers in one file?
Yes. Multispeaker dialogue is generated as one piece of audio: label each line with its speaker, keep the labels consistent, and the engine reads the exchange as a scene rather than as separate clips stitched together.
How should I format a dialogue script?
One line per speaker, a clear label at the start, a line break between speakers, and performance tags in square brackets where the emotion matters. Keep directions out of the spoken text so they are not read aloud.
Can each character have a different voice?
Yes. Assign a voice per character once, then reuse the assignment for every scene in the series. The casting sheet — character, voice, pace, any tags — is the document that keeps a long production consistent.
What is dialogue audio used for?
Audio drama, training scenarios, game chatter, podcast skits, language-learning dialogues and animated shorts. Anywhere two or more people talk and the listener needs to hear who is speaking.
What does a dialogue scene cost?
One credit covers about 1,000 characters of the whole scene, not per speaker. A five-minute two-person scene is roughly 4,500 characters, so about 5 credits; a ten-episode series of the same length lands near 50 credits.
Related
- AI podcast voice - two-voice episode scripts and show consistency.
- AI character voices - casting and directing distinct characters.
- AI voice for gaming - character lines inside a game engine.