Voice Steering

Steering lets you direct how a line is performed — its emotion, pacing, volume, and delivery — using plain-English instructions written directly in your script. It's the fastest way to turn a flat read into a performance without switching voices.

Studio model only. Steering is fully supported on the VocalLab Studio model. On VocalLab Pro and Lite, steering instructions may be read aloud literally instead of performed — switch the model to VocalLab Studio in the generator before using the tags below.

How to write a steering instruction

Wrap your direction in square brackets and place it at the start of the line, before the text to be spoken. No special syntax or numbers — just describe the delivery in your own words.

[say quietly in a low tone with deliberate pauses] We need to talk about what happened last night.

Write the instruction in English even when the spoken text is in another language:

[speak warmly and slowly, like greeting an old friend] Bonjour, je suis ravi de vous rencontrer.

What you can control

You can combine several directions in one instruction:

[say sadly with deliberate pauses in a low voice and a hushed style] I got the call this morning. He's gone.

Non-verbal sounds

Drop these tags inline, exactly where the sound should happen:

[laugh] · [sigh] · [breathe] · [cough] · [yawn] · [clear throat]

That's the funniest thing I've heard all week [laugh] — okay, okay, let's get back to it.

Pauses

For a pause of an exact length, drop a <break /> tag inline, right where the silence should fall:

Let me think <break time="1.5s" /> Yes, I've thought about it.

• Give the time in seconds or milliseconds — <break time="1.5s" /> and <break time="1500ms" /> are identical.
• Up to 20 break tags per request, in every supported language.
• Tag names are case-insensitive, but keep the tag well-formed and self-closing: <break time="1s" />.
• On long scripts, prefer the millisecond form (1500ms) — it's the most robust.
• Break tags are silent markup: they never appear in your captions or SRT files.

For softer, natural pacing you often don't need a tag at all:

• Commas for short beats, periods for a full stop
• Em dashes (—) for a dramatic pause
• A blank line between paragraphs for a longer breath

Building long-form narration? The Audiobook editor also has dedicated pause blocks (0.5–10 seconds, in half-second steps) you can drop between lines.

Emphasis

To stress a specific word, capitalize just that word (not the whole sentence):

We NEED a real vacation.   — stresses “need”
We need a REAL vacation.   — stresses “real”

Best practices

• One instruction set per line, at the very start. Non-verbal tags are the exception — those go inline.
• Keep instructions lowercase and punctuation-free — write [say quietly in a low tone], not [Say Quietly, In A Low Tone!].
• Don't combine conflicting directions (e.g. [whisper] + [very loud]) — the result becomes unpredictable.
• Match the delivery to the words. A cheerful instruction on grim text fights the model and degrades quality.

Examples

[speak as if barely holding back rage, forcing every word through gritted teeth] I have told you. Repeatedly. And you STILL didn't listen.

[overwhelmed with excitement and barely able to contain yourself] We just hit a million users. I still can't believe it — we actually did it!

[articulate clearly at a steady pace] Each step must be followed in order. Do not skip ahead.