Sibelius edition.

What Cantai does

Cantai reads your score (notes, lyrics, dynamics, tempo, expression markings) and generates a realistic sung vocal performance. It works inside Sibelius as a playback device, so your vocal parts play back through Cantai automatically alongside the rest of your ensemble.

System requirements

• Sibelius 2026.6 or later (2026.9 or later to choose every singer by name in the Mixer)
• macOS: Apple Silicon (M1 or later) on macOS 14 Sonoma or later; Intel on macOS 13.4 Ventura or later (renders on the CPU, slower than Apple Silicon)
• Windows: Windows 10 (2020 update) or later, x86-64 processor
• 8 GB of RAM minimum, 16 GB recommended · 4 GB of free disk space

The full list, and what happens on a computer that doesn't qualify, is on System requirements.

Installation & activation

1. Create a free account at cantai.app (or log in if you already have one).
2. Quit Sibelius. Go to Downloads and download the installer for your OS, then run it. If Cantai is already installed, the new installer updates it; your projects and settings stay as they are.
3. Open the account menu (top right of cantai.app) and choose Profile. Your activation token is the Account Token there.
4. Launch Sibelius and open a Cantai window (see below). Paste the token and click Activate. You do this once per computer.

Setting up a vocal part

1. Add lyrics to your score

Select the first note in your vocal part and press Cmd+L (macOS) or Ctrl+L (Windows) to begin entering lyrics. Type your lyrics, pressing Space to advance after a complete word, or - (hyphen) to advance after each syllable of a multi-syllable word.

2. Load & activate Cantai in Playback Devices

Open Play > Setup > Playback Devices. In the Active Devices tab, find Cantai in the Available devices column, select it, and click Activate >> to move it into the Active devices list. Add one Cantai instance per vocal staff you want Cantai to sing; each can sit alongside other devices such as NotePerformer.

3. Assign each instance to a staff in the Mixer

Open the Mixer (M). On each vocal staff's strip, open the input-device dropdown and select the Cantai instance you want that staff to play through.

4. Choose a voice

On the staff's strip in the Mixer, open the sound dropdown:

• Sibelius 2026.9 or later lists every Cantai singer by name, with their voice type (for example Soprano: Aurelia or Tenor: Isaac), plus the choirs and Ottetto ensembles. Pick the singer there, and the staff keeps them when you reorder or rename staves.
• Earlier versions list the voice types (Soprano, Alto, Tenor, Baritone, Bass), the choirs and Ottetto ensembles. Pick a type there, then choose the singer in the Cantai window (Choosing a voice).

A new staff starts with a voice of its type, so it sings even before you choose. The installer sets up the list that matches the Sibelius it finds; if you update Sibelius to 2026.9 later, run the Cantai installer again and restart Sibelius.

5. Press Play

That's it. Whenever you change anything that affects a vocal part (notes, lyrics, dynamics, tempo or expression), Cantai re-renders just the affected segments. No manual refresh needed. If you press play before a part is ready, the built-in synth plays the notes that aren't sung yet, and the singer comes in as each piece arrives.

The Cantai window

Open it with the gear icon on the staff's strip in the Mixer. The first time, it may ask which part of your score this instance sings. After that it puts the singer on a stage: their portrait, their name, the part they sing and the language. While Cantai renders, a light breathes on the portrait and Rendering… shows in the bar.

• The singer's name opens the voice list.
• The language under the name changes the score language.
• The gear, top right, opens Settings.

When Cantai can't sing a part, the window says why, in plain words: the engine isn't running, Cantai isn't activated, or a render failed, with what to do next.

Choosing a voice

You can also choose the singer in the Cantai window: click the singer's name to open Voices. Filter by type on the left, or search by name. Press next to any singer to hear a sample on your computer's speakers, then Choose. The pick is saved with the score, on its staff.

Classical phrasing

Every solo voice can sing with Classical phrasing: the same voice, with the timing and pitch of a classical ensemble singer. The switch appears under the singer you choose; it's on when you pick a new voice. While it's on, the window reads “sings Soprano · Classical phrasing”. It's saved per staff, with the voice. Scores from before 4.0 keep it off until you change it.

Choirs & ensembles

• Choral Female and Choir Male render a whole choral section. When one of them is chosen, a Section size slider appears at the bottom of the window: 3 to 10 voices per part.
• Ottetto ensembles (Sopranos, Altos, Tenors, Basses, Women's Voices, Men's Voices and Full Ensemble) are the eight Ottetto singers, two or more to a line, each in their own seat.

Languages

Cantai sings eight languages. Those marked Experimental in the menu are new and still being refined with native singers.

Choose it with the language under the singer's name. It isn't detected from your lyrics, and it applies to every part in the score: Cantai asks before switching, then renders them again. Spell your lyrics correctly: Cantai's phoneme engine derives pronunciation directly from the text.

Settings

The gear opens Settings. The sections that shape the sound are Playback and Pronunciation; the rest is there when you need it.

• Playback › Cantai level sets the level of every Cantai part in the score at once, from −24 to +6 dB. It's a mix setting: moving it never re-renders.
• Playback › Vibrato goes from straight (0%) to wide (200%) for every Cantai part in the score. Natural is each voice's own vibrato; changing it re-sings the score.
• Playback › Dynamics: with Follow Sibelius, loudness follows Sibelius's playback of your dynamics, like your other instruments. Fixed levels brings back Cantai's own level for each marking.
• Pronunciation: see below.
• Cache › Clear caches… clears rendered audio for this part, this session, or everything. Cleared audio is rendered again.
• Graphics card: Cantai uses your graphics card when it can, so the singing is ready sooner. Turn it off if it causes trouble on your computer.
• Support: write to us or export your logs (see Getting help).
• Reset instance settings…, at the bottom, forgets the part chosen in this window. The voice stays.

Pronunciation

Settings › Pronunciation shows the lyrics of the part. Click a word and write it the way it should sound (for example BLESS-id), or switch to IPA. Then choose where it applies: This one (only that word, in this part), This score (every time it appears, all parts) or All scores. Only that word changes; every other word is sung as before. Changes lists everything you've set and where it applies.

Dynamics & expression

You don't need to draw any MIDI automation. Write your score the way you would for human performers, and Cantai interprets it: written markings shape the voice itself (more effort and a brighter tone at forte), and loudness follows Sibelius's dynamics playback, like your other instruments.

Dynamic markings     p, mp, mf, f, ff, etc.
Hairpins             crescendo, decrescendo
Tempo markings       Allegro, Andante, ritardando, etc.

Tips for best results

• Write singable lines: Cantai performs best on parts a real singer could manage.
• Avoid extremely fast melismatic passages on unstressed syllables.
• Hyphenate multi-syllable words so each syllable aligns to its note. Cantai reads Sibelius's lyric hyphens to map phonemes correctly; while entering lyrics, press - after each syllable to advance to the next note. For example: beau- ti- ful · hal- le- lu- jah · a- maz- ing
• If a word sounds unexpected, check the spelling first: misspelled lyrics produce incorrect phonemes. If it's spelled right and still not how you want it, set it in Pronunciation.

Exporting audio

1. Choose File > Export > Audio.
2. Select the output folder and the staves you want to include.
3. Choose your format: WAV for highest quality, MP3 for a smaller file.
4. Click Export.

Cantai audio exports at full quality: the same engine used for playback renders the final file.

If a staff doesn't sing

Open that staff's Cantai window: it says what's wrong and what to do. If it says Cantai can't find the part (after you renamed or removed a staff in Sibelius), open Settings, choose Reset instance settings…, then choose the part again.

Getting help

If something isn't working, open Settings › Support in the Cantai window and choose Write to support…. Tell us what happened; your logs go with your message, even when the engine won't start.

You can also reach us through the Help chat button in the bottom-right corner of any page on cantai.app. Choose Export logs… in Settings › Support and drag the zip it saves (for example cantai-logs-2026-10-28-085622.zip) into the chat window.

Licensing

One license covers Dorico and Sibelius: $9.99 a month, $99 a year, or $299 for a lifetime license with free updates. The free trial is the full plugin, up to 30 seconds of audio.

Try for freeSee plans

Cantai is made by Turing Opera Workshop. We build tools for composers who take vocal writing seriously.