Is there a way to distort a sound so that can't be restored to its original form (i.e. the function moves one way), but it doesn't add noise to the sound (you can still decipher certain features of the sound based on listening to it; no randomness is added).

  • Are you talking about doing it for disguising a voice like witness protection? Commented Aug 25, 2018 at 7:55
  • For the purpose of witness protection, I highly recommend just re-recording what was said and using a voice actor or a Text-to-Speech system like Google's Cloud API system. cloud.google.com/text-to-speech Commented Aug 29, 2018 at 12:27

1 Answer 1


You should investigate a linear predictive coding mechanism like Silk or Speex. Encode the speech with said codec then decimate the bitrate down to the minimum possible intelligible level. I defy anyone to reverse the processing.

  • Decimating the bitrate is a great idea!!!!!! I'll look into the other ones. Commented Aug 29, 2018 at 17:22
  • using specific Speech codec is the key - the way it works is that it detects vocal inflections, tonal and noise components and then recreates the voice using parametric synthesis. The voice you hear through these codecs isn't actually real it's synthesised on a block-by-block basis.
    – Mark
    Commented Aug 30, 2018 at 5:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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