Help

Everything you need to make a voiceover in Stentor, from the first launch to fixing a word that came out wrong.

Getting started

The first time you open Stentor, it walks you through setup:

Pick what to download. The basic setup (Chatterbox Turbo and Whisper, 5.3 GB) voices English in your voice. Tick extras only if you'll use them.
Download. The progress shows live. Recording and voicing open once it's done.
Record your voice. Say the consent line, then read the text shown, about a minute, in a quiet room.
Voice your first script. A short one is ready for you; it opens in the editor and starts voicing.

Writing with symbols

Symbols in square brackets change how the next words are said. Type @ anywhere in the script to pick one from a list, or type it by hand, like [pause 1]. A mood or pace lasts until the next one, [normal], or the end of the paragraph.

Symbol	What it does
[narration]	Storyteller voice, for telling and explaining
[happy]	Bright and smiling
[surprised]	Raised and curious: a twist, an unexpected fact
[dramatic]	Weighty: a reveal, a big moment
[whispering]	A whisper made from your voice, or your own whisper if you recorded one
[angry]	A hard edge
[sarcastic]	Dry
[advertisement]	Upbeat promo voice
[energetic]	Big energy, with the Standard model (slower to voice)
[calm]	Soft and relaxed, with the Standard model (slower to voice)
[slow]	10% slower
[fast]	10% faster
[normal]	Ends a mood or pace before the paragraph ends
[pause 0.5]	Half a second of silence
[pause 1]	One second of silence
[pause 2]	Two seconds of silence

Sounds: [laugh] [chuckle] [sigh] [sniff] [gasp] [clear throat] [cough] [groan] [shush]. They're made by the voice model as it speaks.

Not sure where symbols go? Enhance adds them for you in a style you choose (see writing assistants), and you see every change before keeping it.

Voicing and the word check

Press Voice it. Stentor voices the script part by part on your Mac, then Whisper listens back to every part. A word that came out wrong is voiced again automatically; anything still unsure is underlined. When it's clean, the status line says every word heard as written.

Edit and voice again: only the changed parts are voiced again. They're tinted, and the button says how many: Voice 3 changes.
Play from a word: click a word while it plays, or select it and choose Play from here.
Earlier voiceovers are kept, and every round of editing is a version you can go back to.
Delivery, in the side panel: how much each take varies, the pauses between sentences and paragraphs, and loudness (−14 LUFS, YouTube's level).

Fixing a word

Select the word and use the toolbar that opens:

Redo this part: voice just that part again. Every take is a little different.
Fix how it's said: your writing assistant works out how to pronounce a name or term, and adds a rule.
Say it like this: write the rule yourself, for every script, like RAG → rag or npm → N P M.

Numbers and names are often flagged even when they're said right: listen to the part before changing anything.

Which models to download

Model	Download it if…
Chatterbox Turbo + Whisper	Always: the basic setup, your voice in English and the word check (5.3 GB).
Chatterbox Standard	You want stronger moods, and [energetic] and [calm] (3 GB; about 5× slower to voice). Without it, moods are lighter and those two are read in your normal voice.
Resemble Enhance	Your recordings are noisy, or you want AI clarity on the finished voiceover (680 MB).
Maya2 Native	You make videos in Hindi, Tamil or 9 more Indian languages (5 GB).
Chatterbox Multilingual	You make videos in Spanish, Arabic, French, German or 17 more languages (3 GB).

Add or remove any of them later in Integrations › Voice models. Removing one frees its space; your scripts and voices stay.

Your voice

In Voices, choose Clone my voice. You say a consent line with a fresh code, and Stentor checks it's the same voice that reads the text. You can clone several voices (calm, energetic), or another person's voice if they're there to say their own consent. You can also record your own whisper, so [whispering] sounds like you.

Languages and translation

Language, in the side panel, is found from the script's text (Auto), or you set it yourself. The Model menu only offers models that speak it.
Mixed scripts: each sentence is voiced in its own language. The language changes at the end of a sentence, so a sentence that mixes two languages is read in one of them. Moods and sounds work in English parts only; with your own voice, each language uses its own model, so your voice can sound a little different between them.
Translate, next to Enhance, writes the whole script in another language as a new script. Select a sentence to translate just that, in place.

Claude Code, Codex and ElevenLabs

All optional, all in Integrations:

Claude Code or Codex power Enhance, Fix how it's said and Translate. Install one and sign in with your own Claude or ChatGPT plan; Stentor finds it. No API key needed.
ElevenLabs: paste your API key to use your ElevenLabs voices and models in the editor. It uses your credits, and the cost shows before voicing.

Exporting

Export saves a folder named after the script, by default in Movies › Stentor: the voiceover (WAV or MP3), SRT captions timed to your voice, word timings (JSON) for animations, and the script as text.

When something goes wrong

Recording is silent

Allow microphone access: System Settings › Privacy & Security › Microphone, turn on Stentor, then record again.

Not enough space

Setup shows how much each model needs and how much is free. Untick extras, free up space, or remove models you don't use in Integrations › Voice models.

A download stopped

Check your connection and choose Try again: what's already downloaded is kept.

A word keeps coming out wrong

Try Redo this part, then Fix how it's said or a Say it like this rule. Spelling it the way it sounds in the script also works.

A sound or mood was skipped (ElevenLabs)

ElevenLabs treats moods and sounds as suggestions. Redo that part, or put the sound at the start of its own short sentence.

Questions

Is there a cost per voiceover?

Not on your Mac: voice the same script as often as you like. ElevenLabs, Claude and ChatGPT plans are your own, and billed by them.

Can I use my voiceovers in monetised videos?

Yes, they're yours, within the terms (above all: only voices you have consent for) and the licences of the models you use.

Does it work offline?

Yes, once the models are downloaded. Writing assistants and ElevenLabs need the internet.

Where are my files?

In ~/Library/Application Support/Stentor on your Mac. Exports go to Movies › Stentor unless you choose another folder.

Still stuck? Write to support@getstentor.com and say what you tried; a screenshot helps.