What is the easiest way to change a voice recording such that the identity of the speaker can't be determined.

The transformation has to be non-invertible. For instance, "pitch changing" won't work because it can be undone.

(I have no experience with audio processing and I'm probably using incorrect terminology. Sorry about that)

  • 2
    A vocoder should do it. Maybe frequency modulation or ring modulation with a randomly varying carrier but that might make it randomly unintelligible. Commented Dec 15, 2015 at 3:19
  • 3
    The word for today is "Random".
    – n00dles
    Commented Dec 15, 2015 at 5:51
  • The real question and answer is, "Why?" Commented Feb 11, 2016 at 14:45
  • 3
    There are a number of legitimate reasons why a person might want to conceal their identity. If someone wants to conduct an interview, for example, without divulging their identity, then obfuscating their voice would be necessary.
    – Dr. Funk
    Commented Feb 11, 2016 at 20:00
  • Mathematician here: As long as you use any "non-invertible" process (e.g., frequency changing), you cannot get the unique original sound. However, the identity of the speaker may still be intelligible. Furthermore, if someone has multiple recordings, or some background information of the speaker (e.g., a guess of the expected frequency range, if one knows their gender), one can combine these pieces of information to get a pretty good approximation of the original sound.
    – zim
    Commented May 12, 2023 at 8:06

3 Answers 3


So long as you're doing some sort of transformation on the vocal sound, it can be reversed.

A vocoder is probably the the best way to go if you need real-time disguising.

If concealing vocal identity is paramount, then you shouldn't be utilizing a voice recording at all. You should instead use a method that totally obfuscates the original vocal signal, and the only way to do that is by using speech-to-text and text-to-speech tools.


some pitch detunes and filters tied to the most random slow and medium fast LFO are very difficult to decode, like i think it would take a typical audio pro a lifetime before he gave up on it, any FX like volume, overdrive, vowel morph, vocoder, just has to be on super random LFO's.


The best thing I know is to use TRAX software from Flux / IRCAM, for which an astonishing video is posted here. The software enables independant Pitch / Formant / Gender / Age control as well as mapping to subtle LFOs, and ability to add further Formant/Spectral envelopes. The software is designed for both speech and singing voice signals.

The easiest way however is actually just to use Pitch Shifting, which is the most available type of audio processing. Although you might know what the original Pitch shift setting was, doesn't mean it is easily reversible or recognisable from another's perspective.

A combination of EQ colouring between 400hz-4khz (Peak/Notch filters) and then pitch shifting could present a more unique result.

Frequency Shifting and Ring Modulation are very limited and can drastically affect the speech intelligibility.

You could try Vocoding software, although the idea of Vocoding is to really use the formant information already present in the voice signal and substitute the 'vocal cords' for another type of vibration (e.g. another source, saw/square wave). This really transforms the sound a lot more than just the 'identity' that you might be referring to, since the identity of one's voice is a product of the entire physical/vocal system, basically all dimensions of one's head, neck, throat, mouth, larynx and vocal cords, etc, which affect the sound a lot more subtly.

  • 1
    I wouldn't trust any combination of only fixed EQ and pitch shift, since both are invertible. I reckon with statistical tools, transient deconvolution etc. it would be possible to infer the settings from a recording. Of course, if the effects are subtly and randomly varied in time, that makes it much safer. Nevertheless, I would rank nonlinear approaches like ring modulation higher (of course not just ring modulation with a fixed oscillator!), since these are in principle much more inversion-resistant, even if set up so as to be still well intelligible. Commented Jan 12, 2016 at 10:43
  • Have you tried TRAX? The Pitch/EQ are both invertible but are the easiest ways to really modify a voice tone. I have used Ring modulation before on a voice, it worked well to obscure the identity and pitch, with only a slight detuning, but also affected the intelligibility of speech consonants, which is why I wouldn't recommend it. What ring modulation are you suggesting? Commented Jan 12, 2016 at 21:43

Not the answer you're looking for? Browse other questions tagged or ask your own question.