Voice changer
Formant shift vs pitch shift
Every voice effect that works is turning two dials, not one. Pitch shift moves the note your voice sits on. Formant shift moves the resonances that tell an ear how big the speaker is. Confuse the two and you get a cartoon; move them together, on purpose, and you get a character.
Formant shift and pitch shift get used as if they were synonyms, and they are not even neighbors. They act on two different parts of your voice, they fail in two different ways, and every convincing character voice you have ever heard is the two of them moving in agreement. If formants themselves are brand new to you, start with what is formant shifting and come back; this page is the comparison.
Pitch shift moves the note
Your vocal folds open and close at some rate while you speak, roughly 85 to 255 times per second for most adults, and that rate is the fundamental: the note your voice sits on. Everything else in the sound stacks on top of it as harmonics, evenly spaced copies at two times, three times, four times that frequency, a comb running up the spectrum.
A pitch shifter moves that number. Shift down and the comb tightens toward the bass; shift up and it spreads higher. The move is measured in semitones, the same steps as a piano, and a good pitch shifter changes nothing else: same words, same pacing, and, crucially, the same apparent size of speaker. The note moves. The instrument playing it should not.
Formant shift moves the instrument
After the folds make the buzz, the sound passes through your throat, mouth, and nasal passages, and those cavities resonate at particular frequencies called formants. Formants are the fingerprint of the vocal tract itself: a long, large tract resonates low, a short, small one resonates high, and your ear reads that resonance pattern, instantly and involuntarily, as the size of the person talking.
A formant shifter slides those resonances without touching the note. The comb of harmonics stays exactly where it was, so the pitch is unchanged, but the envelope that shapes their loudness moves, and with it the apparent body. Same note, bigger instrument. Same note, smaller instrument. If pitch is which key the piano plays, formant is whether the piano is an upright or a nine-foot grand.
What each dial sounds like alone
Formant alone is the easier one to imagine: the same note coming out of a different body. Slide formants down and your exact pitch suddenly sounds like it is leaving a much larger chest; slide them up and the same sentence turns small and bright. It is a surprisingly usable effect on its own, precisely because the note never moves, so nothing about your delivery feels sped up or slowed down.
Pitch alone is where the famous failure lives. Cheap voice effects shift pitch the way a tape machine does, by speeding the whole signal up, and that drags the formants along with the note: push up an octave and the resonances of your vocal tract climb with it, which your ear decodes as a person shrinking to the size of a squirrel. That is the chipmunk, and it is not a pitch problem, it is a formant problem. Even a clean, formant-preserving pitch move breaks down eventually: drop the note far enough while the body stays yours and the two properties stop telling the same story.
Together: a coordinate, not a compromise
A convincing character is the two dials moved deliberately, in proportion. Deeper is pitch down plus formants down, so the voice sits lower and the instrument grows to match. Brighter is pitch up plus formants up: a higher note from a genuinely smaller speaker, not a sped-up recording of you. Every good preset, the giant, the sprite, the low announcer, is a chosen coordinate on that two-dial map, and the craft is in the proportion: the formant move is usually smaller than the pitch move, just enough that the body agrees with the note.
This is why patchd's Pitch Shifter puts both dials side by side in its Beginner group: Pitch spans two octaves in either direction, Formant spans one, and the app's own tooltip says the quiet part out loud, pitch-down plus formant-down together give a convincingly deep voice. How those coordinates map onto deeper and higher characters, and where the dials run out, is the whole subject of deep and female voice changers, explained.
One formula, two targets
Both dials speak semitones, and both cash them in through the same exchange rate:
factor = 2^(st / 12)Twelve semitones doubles or halves a frequency. A pitch shift multiplies the fundamental by the factor; a formant shift multiplies the resonance frequencies of the envelope by it instead. Same arithmetic, aimed at two different parts of the spectrum.
Worked example, the deep move from the figures above. A voice with a 120 Hz fundamental, pitch shifted -7 st, lands at 120 x 2^(-7/12), or about 80 Hz: the note drops a fifth. The same -7 st on the formant dial instead multiplies the resonances: a first formant near 500 Hz slides to about 334 Hz and a second near 1.5 kHz lands around 1 kHz, while the fundamental stays parked at 120. The engine applies exactly this 2^(st/12) factor to the spectral envelope, which is why the two dials cannot step on each other: one owns the comb, the other owns the curve that shapes it.
The engineering difference between the dials is worth knowing too. Formant shifting separates the buzz from the filter, the classic source-filter model, moves only the filter, and re-synthesizes the result, so patchd's Formant Shifter adds no latency of its own. Pitch shifting has to watch enough of the waveform to know what the note even is before it can move it, so the Pitch Shifter exposes a Quality control that trades a short analysis window against fidelity, including a dedicated Deep voice mode whose longer window resolves the tightly packed harmonics of low voices.
Two dials, two nodes, one rack
In patchd the comparison is not theoretical: both dials ship as separate free-form nodes in the rack's voice-fx section, on Studio. The Pitch Shifter carries Pitch and Formant together for coordinated character moves, and the Formant Shifter is the envelope move alone, a low-latency node for changing the body while the note stays yours. Persona, patchd's voice changer, builds its instant DSP voices from exactly these moves, tuned into original characters rather than imitations of anyone real, each with its own keybind and a clickless crossfade between them.
And when the dials are not enough, when the goal is a genuinely different person rather than a resized you, that is the point where DSP hands off to AI conversion, a different trade with a real delay cost, covered honestly in DSP vs AI voice changing. patchd and Persona are in development now. Join the waitlist and we will tell you the moment they are ready.