Gear & Plugins › How-to

Directing AI Vocals in ACE Studio: Breath, Phoneme, Dynamic and the Advanced Parameters

September 23, 2026
This article contains affiliate advertising (PR).

The first thing that bothers you about an AI vocal usually isn't the sound quality. It's that the performance is off. The pitch is right, the words are right, and yet the breath lands in a strange place, the chorus doesn't lift, the ends of lines fall flat.

ACE Studio puts that part under your hands. The premise is that the AI supplies the voice and you direct the performance — so the parameter panel under the Clip Editor lets you move breaths, retime phonemes, shape intensity and adjust the voice itself.

Worth noting: this panel changed during 2026. Power, Soft, Breathy and Chest were folded into a single Dynamic curve. What follows describes the current V2 behaviour, checked against the official documentation on September 23, 2026.

Three places to intervene

The AI parameter panel sits at the bottom of the Clip Editor. On V2 models it splits into three tabs.

The AI parameter panel at the bottom of the Clip Editor (V2 models) Breath & Phoneme Place breaths, set their length Stretch or tighten each phoneme = articulation and timing Dynamic One curve for intensity Only what you draw goes Manual = the shape of the performance enough on its own for most songs Advanced Air / Falsetto Tension / Energy = the voice itself real 0–1 values, handle gently
Fig. 1: Three places to intervene, left to right: how it pronounces, how it swells, and what the voice itself sounds like. Some singers add a Rap or Opera tab.

That order also works as a workflow. Handle how the words come out (Breath & Phoneme), then how the song rises and falls (Dynamic), then reach for the voice itself (Advanced) only where the first two weren't enough. The documentation is explicit that Dynamic is all you need for most songs. Starting with the four Advanced parameters is usually the long way round.

Breath & Phoneme: where it breathes, and how it articulates

A breath before a phrase is most of what makes a human vocal sound human. This tab covers that, plus the length of every consonant and vowel.

The Breath & Phoneme lane BR. h o l d Drag the boundary: one side grows, its neighbour shrinks, the note keeps its length A longer consonant gives the syllable a harder, more deliberate attack A shorter one smooths the transition between notes Phonemes you adjust turn blue and keep their duration through lyric and note changes
Fig. 2: Notes are split into phonemes. You move the boundary between two of them, not the phoneme itself.

Add every breath, then delete the ones you don't want

Add all on the left of the lane drops breaths across the whole clip, placed where ACE Studio judges they belong — typically at the start of each phrase. Adding them all and removing the surplus is faster than placing them one at a time. Drag the left edge of a block to lengthen or shorten a breath.

Retime by moving the boundary

Notes appear split into phonemes. To change a duration you drag the boundary between two phonemes, not the phoneme itself — one side lengthens while its neighbour shrinks, so the note keeps its length.

The effect is predictable: a longer consonant gives the syllable a harder, more deliberate attack; a shorter one smooths the move into the next note.

From the documentation
  • The boundary at the start of a note is fixed here. To change where a note begins, move or resize the note in the piano roll
  • Phonemes you adjust turn blue and lock — they survive changes to notes, lyrics and even the AI voice
  • Stretch or squeeze too far and the model may not follow the new duration precisely

Dynamic: intensity as a single curve

Dynamic governs how hard the voice is working, from soft and gentle to full and powerful. Push it up for choruses and dramatic moments, pull it down for verses and intimate passages.

Dynamic: only the part you draw becomes Manual neutral Auto Manual (what you drew) Auto Left alone, Dynamic is Auto: the AI infers intensity from melody, lyrics and the singer's style Right-drag to erase and that region goes back to Auto
Fig. 3: One curve. Push it up for power, pull it down for delicacy. Brush for freehand, line tool for clean crescendos.

This is what changed. On V2 models, Power, Soft, Breathy and Chest became this one curve. Instead of balancing four controls that each moved the needle a little, you draw intensity and ACE Studio works out the combination of vocal qualities behind it. Open an older project with edits to any of the four and the original lanes stay put, so nothing you already did is lost.

Leave it alone and Dynamic runs in Auto, inferring from melody, lyrics and the singer's style. Only the region you draw switches to Manual, which means you never have to draw the whole clip — fix the parts that bother you and leave the rest to the AI. Right-drag to erase and that region returns to Auto.

Two tools: a brush for freehand, and a line tool for clean crescendos and decrescendos. Some singers add a dedicated tab next to Dynamic, such as Rap or Opera.

Advanced: Air, Falsetto, Tension, Energy

Expand Advanced and four more parameters appear. Unlike Dynamic these aren't envelopes — they're the real synthesis values, 0 to 1, that get fed into the voice. Like AI pitch, they're generated from the singer's style, the melody, the lyrics and the pitch.

The four Advanced parameters (real values, 0 to 1) Air Raspier and brighter as it goes up A slight lift works well on powerful lines Falsetto Does less than you expect on low notes Lower Tension at the same time Tension Low is soft, high is tight and strained The one that unlocks the other two Energy Power — not simply loudness For a gentle sound, drop Tension too
Fig. 4: These four are not independent. The docs pair them explicitly: lower Tension for more falsetto, lower Tension again for a gentler tone.

The important part is that they interact. The documentation pairs them directly: lower Tension if you want more falsetto, lower Tension again if you want a gentler sound out of Energy. When one parameter refuses to do what you expect, its partner is usually holding it back.

Before you start moving these
  • AI-generated values are regenerated easily. Change notes, lyrics or pitch and the parameters for that section are rebuilt — even one changed lyric regenerates every parameter in that piece
  • Anything you edit by hand becomes a user parameter and is left alone
  • To lock the AI's own curve without changing its values, use the fixed brush. The eraser removes your edits and restores the generated values
  • The docs repeat the same advice throughout: keep adjustments moderate. Extreme values produce unexpected performances
ACE Studio — the AI sings, you direct
160+ royalty-free AI voices and 40+ AI instruments, unlimited. Rent-to-own: two years of subscription converts into a permanent licence.
Check out ACE Studio →
Pricing and specs change; see our ACE Studio review and the official site.

How far to take it

In practice the priority order is clear.

  1. Breaths first (Add all). The single biggest step toward sounding human, and the cheapest
  2. Then Dynamic. Lift the chorus, drop the quiet parts. The shape of the song is mostly decided here
  3. Then individual phonemes for the specific words that come out mushy, or too hard
  4. Advanced last, and only when you can say out loud what you want — "this one line needs more air", not "let's see what this does"

Which is the same thing as saying that opening with the Advanced parameters is inefficient. There's probably a reason the documentation keeps repeating that Dynamic is enough for most songs.

The official tutorial (with video demos)

ACE Studio publishes a tutorial that demonstrates these moves in short videos: converting a rough vocal to MIDI, adding a vocal run, editing breaths and phonemes, adjusting Dynamic and the expression controls, adding doubles and harmonies, and before-and-after comparisons — 11 videos across six chapters.

These are the kind of edits that are much easier to watch than to read about, so it's worth going through before you open the panel yourself.

ACE Studio AI Voice Editing Tutorial (official) — Vocal-to-MIDI, breath, phoneme, Dynamic, Air / Falsetto / Tension / Energy, doubles and harmonies, in 11 demos

Frequently asked questions

Q. What can I actually change after the AI sings?

A. On V2 models the AI parameter panel has three tabs: Breath & Phoneme, Dynamic and Advanced. You can place breaths and set their length, adjust the duration of individual phonemes, draw an intensity envelope across a phrase, and fine-tune Air, Falsetto, Tension and Energy. Some singers add an extra tab such as Rap or Opera.

Q. Where did Power, Soft, Breathy and Chest go?

A. On V2 models those four were merged into a single Dynamic curve. Rather than balancing four controls that each had a subtle effect, you draw one intensity curve and ACE Studio translates it into the right combination of vocal qualities. Projects that already contain edits to any of the four keep showing the original four lanes so existing work is preserved.

Q. Should I always use the Advanced parameters?

A. The official documentation says Dynamic is all you need for most songs. Air, Falsetto, Tension and Energy are for precise control over one specific vocal quality, and the docs advise keeping adjustments moderate, since extreme values can produce unexpected performances.

Q. Do my manual edits survive a lyric change?

A. Yes. Anything you edit by hand becomes a user parameter and is not regenerated when you change notes, lyrics or pitch. Phoneme durations you adjust turn blue and lock. Sections left on AI-generated values behave the opposite way: changing even a single lyric regenerates every parameter in that piece.

Summary

When an AI vocal sounds wrong, it's usually the performance rather than the pitch — and in ACE Studio the performance is editable.

You don't need to shape all of it. Adding breaths and lifting the chorus with Dynamic already changes the impression considerably. Pick one phrase and move it between Auto and Manual a few times — that's the whole idea in miniature.

This article contains affiliate advertising. Specs and pricing are as confirmed on the official documentation and website on September 23, 2026. Check the official pages for the latest.