How Do Voice Changers Work?

Your voice is a buzz from your vocal cords, shaped by your throat, mouth and nose. A voice changer can't reach into your throat, but it can change the sound wave on its way out. Nearly every effect you've heard is a mix of three tricks.

What does a voice changer change?

The trick is moving pitch and formants separately. If you only raise the pitch, you get a chipmunk: the old sped-up tape sound. Raise the formants with it, a little less, and you get something that sounds like a different person instead of a cartoon. You can hear the difference on your own voice with our online voice changer: record a line, pick the Girl voice, then move only the pitch slider.

What does that look like?

A spectrogram is a picture of sound. Time runs left to right, low notes are at the bottom, high notes at the top, and brighter means louder. Here's one sentence, "Hey, can everyone hear me okay?", before and after three of our presets:

• Girl raises the pitch 8 semitones and the formants 2. The stripes (the harmonics of the voice) spread out and move up.
• Deep lowers the pitch 7 semitones and the formants 3, then evens out the volume. The stripes squeeze together near the bottom.
• Robot doesn't change the pitch at all. It multiplies the voice by a 55 Hz tone (that's the metallic buzz), adds a resonant flanger and crunches the sound to 10-bit. You can see the voice smeared into vertical bands.

How does a voice changer work live, in a call?

A live voice changer does the same thing in tiny slices, many times a second. It takes a few milliseconds of your mic, changes it, and passes it on while the next slice arrives. That delay has to stay small or you'll talk over people. Our pitch shifter works on windows of about 20 ms, so at 48 kHz it adds 18.7 ms, plus 1.5 ms for the limiter that stops loud bits from clipping. Effects without a pitch change, like Robot, add almost nothing.

Then the changed sound has to reach your call. Desktop apps like Voicemod add a virtual microphone to your computer that every app can pick. A browser real-time voice changer like ours can only reach websites open in Chrome, through our extension, where it shows up as a mic you pick in Meet, Zoom on the web or discord.com. Our voice changer hub lists which of our tools work where.

How are AI voice changers different?

An AI voice changer doesn't bend your sound wave. It's trained on recordings of a target voice. When you speak, it works out the words, timing and tone of what you said, then generates new audio of the target voice saying it the same way. That's why it can make you sound like a convincingly different person, and also why it needs a strong computer or a server, adds more delay, and sometimes garbles words it didn't catch.

Newer models don't even need training on each voice: they copy a voice from a short reference recording. Live AI voice changers still can't work word by word. They re-speak your voice in short blocks and have to hear a little of what comes next before finishing each one, so you hear yourself a moment late. Our AI Voice Changer avoids that by working on a recording: you record or upload a clip, and it re-speaks the whole thing in the voice you pick.

Why do voice changers sometimes sound robotic?

• Big pitch shifts. The further you move pitch, the more artificial it sounds. Shifts of a few semitones sound natural; a full octave rarely does. Our Girl and Deep presets move 9 and 7 semitones.
• Pitch without formants. Without the formant shift, high voices sound like chipmunks and low ones like slowed tape.
• Noise in your mic. The changer changes the noise too. A fan becomes a weird whine. Fix your mic first (our mic guide covers it).
• Your own voice is far from the preset's target. Presets are tuned for an average adult speaker. Ours has Pitch and Tone sliders so you can fit a preset to your voice.