Voice Disguising

[MIKEROPHONICS]

I have to disguise some interviews on a documentary so that the internees are not executed by the fundamentalist "organisation" they work for.
Politics aside, believe it or not I have not done this in my 30 years of post. Do people just pitch down and flange slightly?

Any suggestions?

[FlorianE]

A few years back, I was in that situation and went there with good success.

Florian

[MIKEROPHONICS]

I used Excalibur - worked a treat

[MIKEROPHONICS]

sorry - didn't means to sound rude. I had a producer hanging around, so needed a solution to be agreed fast.

That does look like an interesting plug-in - I had a quick watch and may follow up on it

[Frank Kruse]

I wonder if anyone has checked if these methods are really forensic-proof methods to protect a critical source in a documentary. I mean the responsibility the audio guy has in that moment is quite high when he treats a voice that needs to protect a source of information from being being traced down by a potential adversary. I know it depends on your threat model but better make sure you know that the process is irreversible when doing it.

Ultimately a voice can also be identified when you know the unique things they do when speaking like fill words "ahhh, umm etc. " in a typical pattern. Some people have such characteristical patterns when they speak you could ID them even when it was a vocoded voice set to 100%.

Which is why often lines are re-read by a different person to really protect the source.

Better keep that in mind when you make such offers and your source could be facing real threats.



Frank.

[FlorianE]

Re-reading certainly is safer than any audio transformation, especially with strong accents, dialects or speech idiosyncrasies.
But it can also destroy the emotional impact of a statement, which can be crucial.
If you must use transformation, then I prefer the 'inconspicuous' approach, to make it sound as natural as possible, to avoid arousing suspicion or attempts to reverse engineer. Plus a bit of discreet editing out of obvious patterns, if possible.

Florian

[Frank Kruse]

Originally Posted by FlorianE

Re-reading certainly is safer than any audio transformation, especially with strong accents, dialects or speech idiosyncrasies.
But it can also destroy the emotional impact of a statement, which can be crucial.
If you must use transformation, then I prefer the 'inconspicuous' approach, to make it sound as natural as possible, to avoid arousing suspicion or attempts to reverse engineer. Plus a bit of discreet editing out of obvious patterns, if possible.

Florian

Well, that's the thing. Would you be able to take responsibility for the safety of a source on the basis of artistic decision-making?
Are you going to tell the source that your director promised anonymity "sorry, they found you and burnt down your house but the film was so much more emotional using your own voice with the treatment our sound-guy suggested"?

source-protection should not be taken lightly.

What I mean is that you should be aware of what you're doing when you tell your director that the process you're using is safe. Depending on the topic and threat model the responsibility can be huge. And once the film is out there, there's no way to undo it.

[FlorianE]

Originally Posted by Frank Kruse

Depending on the topic and threat model the responsibility can be huge.

I completely agree. Safety requirements all depend on the risk involved. And we should never put anybody at risk.
It is a moral issue, not a technical one. Age-old dilemma, and a can of worms.

Florian

[tamasdragon]

I did things like this quite often for a show I mixed. For the reasons mentioned above, I always used audiosuite processing on the voice.
Developed a tried and good method on my own voice, then tried to reverse-engineer it, and when it was working, I just made some plugin presets and used them as audiosuite processing on the real material.

My chain included distortion, eq, pitch, harmonizer, additional flanger. It's always a great deal to keep it intelligible while still have to disguise it so noone can recognise. Back then I wasn't making notes though, so if such work would come, I'd have to reinvent my past methods again.

[minister]

I agree with Frank. Often when I see this done I think, "If I knew this person, I would recognize them." Mostly because of, as was mentioned, speech patterns and sometimes pretty bad job of visual masking where they are just in a shadow. They don't mask enough. I realize that there is a balance between intelligibility and masking, but often times someone's life is at stake.

That said, Trax is a really interesting plugin! I have done some amazing things with it, like age voices! But like most IRCAM stuff, it is super unstable in Pro Tools. I have to print it. If I automate it, it can kill a session upon printing.