
Shift a speaker's pitch by semitones, preview eight seconds, stack two passes, and learn what pitch alone cannot hide.
You want a 90-second clip of a real support call for a sales demo, but the customer on the recording has not agreed to be recognisable. The Audio Scrambler shifts a voice up or down by a set number of semitones and keeps the timing, which makes the speaker harder to place at a casual listen. It is a disguise, and being clear about how thin a disguise it is makes the difference between a sensible use and a risky one.
The short version: set the slider to a large negative value such as -6, press Preview 8 s to hear the first eight seconds, then run Disguise Voice for the whole clip. Before you do any of that, cut out the names and account numbers, because a pitch shift does nothing to the words.
The tool moves pitch by an amount you choose and then compensates the tempo, so the clip stays the same length and the words keep their speed. The slider runs from -8 to +8 semitones in steps of 1. Presets are Strongest Deep (-8), Deep Voice (-6), Slightly Lower (-3), Slightly Higher (+3), High Voice (+6) and Strongest High (+8). At 0 the buttons are disabled, since nothing would change.
Each semitone is a frequency ratio of about 1.059. A speaker whose voice sits around 180 Hz lands near 127 Hz at -6 and near 113 Hz at -8. Going up, +6 takes 180 Hz to about 254 Hz. Lower shifts tend to sound more natural on speech than higher ones, which can turn thin and cartoonish.
This is not encryption and it is easy to reverse. Anyone with the output can shift it back by the opposite amount and get close to the original voice. Cadence, accent, favourite phrases and what the person says can identify them regardless of pitch. The tool's own note calls it a layer, not the only layer.
So use it for low-stakes cases: a demo where the speaker is already comfortable with the recording being shared internally, or a sample where you simply do not want a colleague's voice used publicly. For anything involving real confidentiality, legal exposure or personal safety, do not rely on pitch alone. Get consent, use a transcript, or re-record with an actor.
If a name occurs mid-clip, trim around it, or mute it in your editor. The tool does not detect or bleep words.
The page explains that one shift is easy to undo and suggests running the result through again with a different shift. The sum is what matters: -8 followed by -6 gives a net of -14 semitones, a frequency ratio of about 0.44, which would take 180 Hz down to roughly 80 Hz and may sound very low and slightly processed. Be careful with opposite signs. -8 followed by +6 nets to -2, which undoes most of the disguise.
Listen to the second-pass result at the same volume as the original on a phone speaker, where most viewers will hear it. If words become hard to follow, back off. A disguise that makes the demo unintelligible has defeated its own purpose.
Treating a pitch shift as redaction. It hides tone, not content.
Using a tiny shift such as -1 or +1. At that level the voice remains recognisable to anyone who knows the person.
Choosing a very high shift for a male voice or a very low one for a female voice and expecting it to sound natural. Test both directions on the eight-second preview.
Sharing the original file alongside the disguised one. Keep the source recording in a restricted folder, and only circulate the processed clip.
Forgetting that background sounds are shifted too. A distinctive hold tune or a colleague's voice in the background also changes pitch, but may still be recognisable.
If your presentation tool wants a specific format, convert with the Audio Converter. The processing happens in your browser, so the call audio is not uploaded to a server. Test your settings with Preview 8 s before committing to the full file, since the full run is the one you want to get right the first time.
You can find the rest of the sound utilities in the audio tools category.
What are you trying to disguise next, a single voice or a whole panel? Tell us on the contact page.