Can MetaSound shift vocal formants independently of the fundamental frequency?

danporto · May 3, 2026, 1:00am

I’m looking for a way to shift vocal formants in MetaSound while keeping the fundamental pitch unchanged.

My use case is a planetary atmosphere audio system in Unreal Engine. I want to process voice in a way similar to the helium-voice effect, where the vocal colour/formants change while the fundamental pitch largely remains the same. In other words, I want the voice to sound as if its resonances have shifted, without simply transposing the whole signal.

I know the built-in Pitch Shift node changes pitch and formants together, which is not what I need.

Has anyone found or built a way to do this in MetaSound, either with existing nodes, a filter-bank workaround, or a custom C++ MetaSound DSP node?

I would be interested in any examples, partial implementations, or advice on whether this is realistically achievable inside MetaSound.

danporto · June 7, 2026, 10:19am

I found a workable solution to my own problem, so I thought I would share it in case anyone else is trying to do something similar in MetaSound.

MetaSound does not seem to have a dedicated node for independently shifting vocal formants but I built a useful approximation using a resonant filter-bank approach.

The system does not pitch-shift the source audio inside the formant patch. Instead, it changes the apparent vocal colour by moving a set of voice-like resonances. In other words, the voice can sound darker, brighter, larger or thinner, without this patch deliberately changing the main pitch of the sound.

This is not true formant analysis or resynthesis. It does not listen to the input voice and detect its real formants. Instead, it uses a designed set of Biquad Parametric EQ filters. These filters act like artificial vocal resonances, and their centre frequencies are shifted with a control called VoiceFormantOffsetST.

The basic idea is:

VoiceFormantOffsetST uses semitone values to move the formant filters up or down, but it does not transpose the source audio.

That value is converted into a frequency multiplier using:

2 ^ (VoiceFormantOffsetST / 12)

For example, +12 moves the formant filters up by one octave, -12 moves them down by one octave, and 0 leaves them unchanged.

The multiplier is then applied to a set of base formant frequencies, which feed a serial Biquad filter bank.

The filtered signal is then blended with the dry signal using a crossfade.

I am also considering simple voice-profile presets, such as neutral, lower voice and higher voice. These would not change the pitch. They would only change the base formant-frequency values before the formant offset is applied.

So it is not a fully accurate formant shifter, but it works well as a MetaSound-only approximation for making a voice sound darker, brighter, larger or thinner without using the formant patch to transpose the source audio.

The final screenshot shows how I inserted the patch into the main MetaSound Source after Delay Pitch Shift and before the main atmospheric processing patch.