How do I keep a long read sounding consistent?

Direct once at the top rather than per line, split at paragraph breaks rather than mid-sentence, and fix pronunciations the first time you hear them wrong. Consistency comes from having fewer decisions in the script, not more.

How do I write a script with two people in it?

Assign a voice per speaker and keep it constant. Each speaker's lines are segments with their own voice, and the studio adds a gap between them: 0 to 3 seconds, in steps of 0.1. Default 0.5 seconds.

Why does the silence control only sometimes appear?

Because it only applies once a script actually contains more than one voice. Until then there is no gap between speakers for it to set, so it stays hidden rather than sitting there doing nothing.

When should I use Audiobook Studio instead?

When the work is a manuscript rather than a script. The test is not length — it is whether you need the tool to track which parts are done, keep a cast consistent across chapters, and package the result to a retailer's spec. If you will listen to the whole thing yourself in one sitting, you are still in the right place.

Can I move a script from here into Audiobook Studio?

You can paste the text in, and Audiobook Studio will split it into chapters and paragraphs itself. Directions written for a segment here are not carried over automatically — the audiobook workspace has its own per-paragraph direction, emotion and pause controls that replace them.

Is there a limit on how long a single generation can be?

Yes, and it is a practical one rather than a number worth memorising: past a certain length a script is better as several segments regardless, because each segment can then carry its own direction and be regenerated on its own. If you are asking the question because you have a book, the answer is Audiobook Studio.

Voice Studio

Long form and multiple speakers

Holding one character steady across a long read, writing a script with several voices in it, and knowing the point where Audiobook Studio is the right tool instead.

Lesson 4 · advanced · 8 min read

Open Voice Studio

Everything in the first three lessons works on a line. This one is about what breaks when there are four hundred of them — and what breaks is almost never the individual reads. It is consistency: the same character sounding like two people, a passage that stayed dramatic long after the drama ended, a name pronounced two ways in the same script.

Holding one character steady

Consistency in a long read comes from having FEWER decisions in the script, not more. Every per-line direction is a thing that can disagree with the line above it; a single direction at the top of a passage cannot.

Set the direction once, at the top. A direction carries forward until another one replaces it, so a script with one direction at the start is more consistent than a script with a direction on every line.
Return to neutral explicitly. Because directions carry forward, the way to end an emotional passage is to give the next segment its own plain direction — not to leave the brackets off.
Split at paragraph breaks, not mid-thought. Each segment is generated as a unit, so a break placed inside a sentence is where the seam will be audible.
Keep one voice per character for the whole script. Swapping a character's voice halfway is the single most noticeable inconsistency, and it is the easiest one to introduce by accident when a script grows.
Fix pronunciation with IPA the first time you hear it wrong. A name mispronounced in segment two will be mispronounced in segment forty, and it is one token either way.

Careful

The carry-forward rule is the one that bites at length. A direction stays in force until another replaces it, so an emotional passage keeps colouring everything after it until you explicitly hand the next segment a plain direction. In a short script you notice within seconds; in a long one you find it on the third listen.

More than one voice

A multi-speaker script is segments with different voices assigned to them. The studio adds a gap between speakers, and that gap is the one setting that exists specifically for this case: 0 to 3 seconds, in steps of 0.1. Default 0.5 seconds.

Control: Range
Speed: 0.5× to 2×, in steps of 0.1. Default 1×.
Silence duration: 0 to 3 seconds, in steps of 0.1. Default 0.5 seconds.

The controls that matter in a multi-voice script.

Cast before you write
Pick a voice per character first and keep it. Casting as you go is how a character ends up with two voices, and the fix is regenerating everything they said.
Make them distinguishable, not just different
Two voices of a similar pitch and pace are hard to tell apart in a scene even when they are obviously different in a preview. Audition them back to back, not one at a time.
Set the gap once
Too short and the exchange sounds like one person interrupting themselves; too long and the scene loses its pulse. A default gap is right more often than a tuned one.
Direct per character, not per line
Give each character their standing direction where their first line appears. It carries forward, so a character with a consistent manner needs it stated once.

Where Audiobook Studio takes over

The boundary is not a word count. It is whether you are performing a SCRIPT or producing a MANUSCRIPT — whether you will hear every second of the result yourself, or need the tool to tell you which parts are wrong.

Stay in Voice Studio: Move to Audiobook Studio
A script you can see in one screen or a few — an ad, a narration bed, a scene, a voicemail.: A manuscript with chapters. Audiobook Studio splits it, tracks which paragraphs are narrated, and packages the result to a retailer's spec.
You are still deciding how a line should sound.: You have decided, and now need the same decision applied consistently across hours of audio.
You will listen to the whole thing yourself in one sitting.: You will not — and you need the tool to tell you which paragraphs are wrong rather than finding them by ear.

Which tool the work belongs in.

Audiobook Studio is the same speech engine with a production workflow around it: it splits a manuscript into chapters and paragraphs, tracks which of them are narrated, listens back and flags the ones that swallowed a word, holds a cast across an entire book, and packages the result to the spec a retailer checks. None of that exists here, and none of it is worth having for a thirty-second script. The Audiobook track covers it end to end.

Tip

If you are unsure, the deciding question is whether you will listen to the whole thing. If yes, stay here — you are the quality check, and this tool is faster. If no, you need a tool that does the checking, and that is the other one.

What stays hard

Long-form synthetic speech has failure modes no setting removes, and it is worth knowing them before you commit an afternoon.

Sameness. A voice reading forty minutes at one energy is fatiguing in a way that no individual segment reveals. Vary the direction between sections deliberately.
Seams at segment boundaries. Splitting mid-thought puts a join where the ear expects continuity. Split where a person would breathe.
Names and jargon. These are where a long script goes wrong, and they are rarely in the opening lines — so a first segment that sounds perfect proves less than it feels like it does.
Your own ear. After the fourth listen you stop hearing the read and start hearing the words. Get someone else to listen once before you ship anything long.

Questions

What people ask about this

Free to start

Now go make one

Reading about a style description only gets you so far. The studio is free to use — write one sentence and hear what comes back.

Create a free account

1,000 credits every month. No card required.

All Voice Studio lessons · Producing a whole book instead

Holding one character steady

Set the direction once, at the top. A direction carries forward until another one replaces it, so a script with one direction at the start is more consistent than a script with a direction on every line.

Return to neutral explicitly. Because directions carry forward, the way to end an emotional passage is to give the next segment its own plain direction — not to leave the brackets off.

Split at paragraph breaks, not mid-thought. Each segment is generated as a unit, so a break placed inside a sentence is where the seam will be audible.

Keep one voice per character for the whole script. Swapping a character's voice halfway is the single most noticeable inconsistency, and it is the easiest one to introduce by accident when a script grows.

Fix pronunciation with IPA the first time you hear it wrong. A name mispronounced in segment two will be mispronounced in segment forty, and it is one token either way.

Careful

More than one voice

Control: Range
Speed: 0.5× to 2×, in steps of 0.1. Default 1×.
Silence duration: 0 to 3 seconds, in steps of 0.1. Default 0.5 seconds.

The controls that matter in a multi-voice script.

Cast before you write

Pick a voice per character first and keep it. Casting as you go is how a character ends up with two voices, and the fix is regenerating everything they said.

Make them distinguishable, not just different

Two voices of a similar pitch and pace are hard to tell apart in a scene even when they are obviously different in a preview. Audition them back to back, not one at a time.

Set the gap once

Too short and the exchange sounds like one person interrupting themselves; too long and the scene loses its pulse. A default gap is right more often than a tuned one.

Direct per character, not per line

Give each character their standing direction where their first line appears. It carries forward, so a character with a consistent manner needs it stated once.

Where Audiobook Studio takes over

Stay in Voice Studio: Move to Audiobook Studio
A script you can see in one screen or a few — an ad, a narration bed, a scene, a voicemail.: A manuscript with chapters. Audiobook Studio splits it, tracks which paragraphs are narrated, and packages the result to a retailer's spec.
You are still deciding how a line should sound.: You have decided, and now need the same decision applied consistently across hours of audio.
You will listen to the whole thing yourself in one sitting.: You will not — and you need the tool to tell you which paragraphs are wrong rather than finding them by ear.

Which tool the work belongs in.

Tip

What stays hard

Long-form synthetic speech has failure modes no setting removes, and it is worth knowing them before you commit an afternoon.

Sameness. A voice reading forty minutes at one energy is fatiguing in a way that no individual segment reveals. Vary the direction between sections deliberately.

Seams at segment boundaries. Splitting mid-thought puts a join where the ear expects continuity. Split where a person would breathe.

Names and jargon. These are where a long script goes wrong, and they are rarely in the opening lines — so a first segment that sounds perfect proves less than it feels like it does.

Your own ear. After the fourth listen you stop hearing the read and start hearing the words. Get someone else to listen once before you ship anything long.