When trying to hide the identity of people I interview, I have to hide their identity for their safety. Pitch shifting was recommended but anyone can shift the pitch in opposite direction and that is where I am in need of a way to help overcome this problem. Earlier, we used a voice over but due to covid, the artist who was helping in voice over is not available. Any suggestions would be apreciated.
You can pitch shift both up and down with Adobe Audition.
I think that is part of the problem - @dorkter doesn’t want somebody to pitch the voices back down once they were pitch-shifted up (or other way around). So is there a solution that will not allow somebody to bring the voices back to what they were before?
Pitching down two semi tone and pitching down two semi tones on the same audio (on a separate track) is still far from distorted. If pitched back two semi tones can reveal the identity. Is there a way to vary the amount of pitch shift over time, like an expression of time on timeline?
No is the honest answer. Any audio can be pitched up or down so pitch shifting alone is not the solution.
To disguise the voice you’d have to pitch shift it in combination with something destructive also incorporated which end listeners would not be able to simply reverse.
Exactly. An intentional damage (pitch shift) that could not be reversed, at least not easily and not quickly like an opposite pitch shift.
You could apply one of the pitch shift presets which would make it more difficult to reverse.
It’s already been mentioned that if you pitch shift down 6 semitones someone else can pitch shift back up 6 semitones and get the near original voice back. This method is not good enough if you want to fully disguise a voice.
If you really want a completely anonymous voice then getting a voice actor (preferably one that hasn’t heard the original speaker and is reading from a script) would be the way to go. Note there may be other things that give someone away such as the way they say things, their crutch words (such as, “you know” or “know what I mean”, for example) or even the amount of umms and ahhs.
However, there are a few ways I go about attacking this to try and anoymize a voice as much as possibe.
1. Varied Pitch Shifting
Instead of a straight -6 semitone pitch shift use automation lines in the multitrack of Adobe Audition to vary the pitch shift up and down over time. This could still be undone if someone really knew what they were doing. Perhaps add some chorus with heavy modulation to make things harder. Can you tell this is me?
Here’s a screenshot of what I did in Adobe Audition:
Let’s try pitching this sample back up 6 semitones…
Oh no! I guess it’s on to method 2 for higher security
A vocoder plugin is really hard to undo. The plugin I really like for vocoding a voice is called OVox. Here’s my voice with the Fat Boy preset added. Can you tell it’s me?
3. Text to Speech
A final snip if you can’t use a voice actor just use an online text to speech tool such as Google Translate or Amazon Polly. The latter has better audio quality, allows you to download audio files and is almost free.