Voice filtering for real anonymity

C

custerc

Well-known member
I am considering a documentary project that would entail lots of anonymous interviews. As most of the subjects would be at serious risk of legal and potentially life-threatening consequences for their actions, if I were going to shoot this I would need to do a lot of things to ensure security, but one of them would be to filter their voices in a way that was irreversible and untraceable.

So what tools out there exist for this, and how reliable are they? Is this something that can be done with software in post? Is there something one can attach to a mic or run the XLR cord through that will filter voices while shooting (this would be better, as it would mean even footage was stolen from me before I was able to edit it, the people in the interview would be safe (obviously, I'd be shooting them so their faces aren't lit, or would have them wearing masks or something.

Anyone know anything about this tech?
 
Ok, just had a random idea: After you record their interviews put someone else to reenact and even so distort their voices, state that you did so in the beginning of the documentary.

However, I'm also curious about what you asked, I'm not sure what kind of change would be irreversible.
 
If people are going to be physically at risk if your voice scrambler is reverse engineered, this is a security issue. Whatever method you go with, you had better be sure that somebody smarter than you doesn't know how to reverse what you did and recreate the original voice. For example, if you just pitch the voice down, that would sound good at first listen but would be fairly trivial to change back.

Make absolutely sure you get this detail right. If I were you, I'd find a proven solution by people who have done this before, and take everything you hear on any internet forum with a huge grain of salt.
 
robotsound said:
If people are going to be physically at risk if your voice scrambler is reverse engineered, this is a security issue. Whatever method you go with, you had better be sure that somebody smarter than you doesn't know how to reverse what you did and recreate the original voice. For example, if you just pitch the voice down, that would sound good at first listen but would be fairly trivial to change back.

Make absolutely sure you get this detail right. If I were you, I'd find a proven solution by people who have done this before, and take everything you hear on any internet forum with a huge grain of salt.
Click to expand...


Of course, that's why I'm asking here. But regardless, if we ever go through with this project I will talk to some experts first. Pietro's suggestion seems like the safest idea, frankly. That would be impossible to reverse-engineer, and having actors read real words worked for that movie about the Nanjing massacre (Nanking? i forget the title...)
 
My first instinct would also be to re-record with actors. Just about anything you do that won't totally mangle the audio is at least theoretically reversible.
 
Well just about anything you do to it will be irreversible once you destructively process the file. But I suspect you mean you don't want it to be made anything close to the original.

Pitch shifting is a usual thing for this but one could get close gong backward but if you use some formant changing pitch shifting then it's pretty impossible to get close going back unless you know what your trying to get to. Of course if they know what they are trying for they already know who the voice belongs to. Throw on some modulation and your probably at a place where it's very safe, but your also at a place where you are probably going to need subtitles.

I'm with others, if security is that big a deal you need to replace the voices. Take your hero voices and mangle the heck out of them and then do the VO translation as if they were speaking in a different language. You could do it live - I can point you to some filmmakers who managed unidentifiable dialog, (with out even trying), but you would need to transcribe live. Though you could do essentially a simultaneous translation. Use two mics, one feeds the camera and is muffled and mangled (you could use some stompbox if you cant trash it enough on it's own) the other feeds a pair of headsets and the person wearing the headsets repeats what they hear and that is recorded to your second track. That way you have no recording in the clear and you the translation on tape so you can edit.
 
custerc said:
I am considering a documentary project that would entail lots of anonymous interviews. As most of the subjects would be at serious risk of legal and potentially life-threatening consequences for their actions, if I were going to shoot this I would need to do a lot of things to ensure security, but one of them would be to filter their voices in a way that was irreversible and untraceable.

So what tools out there exist for this, and how reliable are they? Is this something that can be done with software in post? Is there something one can attach to a mic or run the XLR cord through that will filter voices while shooting (this would be better, as it would mean even footage was stolen from me before I was able to edit it, the people in the interview would be safe (obviously, I'd be shooting them so their faces aren't lit, or would have them wearing masks or something.

Anyone know anything about this tech?
Click to expand...

Record the interview subjects' shadows, not their silhouette. This will stop any chance of seeing their identity by people brightening their TVs.

Then dub in a voice actor's voice and make a note of that somewhere for your audience like Pietro mentioned. Try not to record the voice actor in a studio with a voiceover-type mic (it will sound too dry/phoney) record it like you recorded the on-camera interview.
 
You could take 2 instances of the voice file, and pitch one down, and one up, then blend them. That way when someone tries to pitch the voice back to the original, there will still be the other voice in there confusing them.
 
Sounds a good idea but sound really horrible and difficult to understand. Shoot the subjects so you cannot see their mouths and dub it - far safer and much, simpler. If you do it properly, many people won't even notice.
 
You must log in or register to reply here.
Back
Top