Sound Rescue Problem.

The problem is perfectly normal, but I don't think we ever addressed it quite this way.

The client shot a sound track of someone now dead. No reshoot. The track is very noisy.

[www.theban.be]

Yes, he's in Belgium.

The difference is the client wants a clear track back no matter what. Who does this?

I was able to produce the usual gargly track with my sound tools--both television and audio. I got him to send an uncompressed .WAV of the same thing instead of the 10:1 compressed MP3 and that was a little easier to deal with--but not much.

So who does he write the check to?

Koz

Soundtrack Pro should be able to handle this quite well and simply. That, or find Gene Hackman. He did this in The Conversation and it sounded great.



Michael Horton
-------------------

Yes,
Find yourself an FBI or NSA agent. The latest technology in this field is developed on the government dime only later to be integrated into applications available to you and me.

<<<Soundtrack Pro should be able to handle this quite well and simply.>>>

It should, but no. Not in our STP. I get about the same results that I do in our other sound packages. There's something about this noiseprint that's pretty fatal. If you do an analysis of the room tone, it has energy all up and down the spectrum. It's not like multiples of power line or fan whine and harmonics. I know how to get rid of those.

I know where I heard this before. I thought it sounded vaguely familiar. I know it's coming from multiple motor noises and fans on a recording machine, but it's the same sound that they broadcast during the war to keep people from listening to Radio Free Europe or Voice of America.

It was a pain in the butt then, too.

<<<The Conversation>>>

in 1974? What version of Sound Track Pro was he on?

Koz

Hi Koz...

Here's a little work on the file I did on the old version of Adobe Audition (1.5) from my laptop PC I have at home.

www.hd-vo.com/gm2.mp3

Only had about ten minutes available to work on this, but I think with a bit more time, the track may be made useable. I'll be back in the studio on Friday, and will have a few more tools - I'd be interested to see what could be accomplished with Soundtrack Pro.

What I did was capture the noise reduction profile, then applied that to the "noise reduction" process at only 75 percent - which allowed enough original audio to leave it understandable (well probably, since I don't speak the language.) I then applied some amplitude expansion/gating to reduce the apparent noise still further.

I'm pretty sure that if some more time was spent, you'd probably get a usable track. Especially if we start with non-compressed audio.

Travis
VoiceOver Guy and Entertainment Technology Enthusiast
[www.VOTalent.com]

Why don't you try posting the sample on the ProTools forum? (http://duc.digidesign.com/)

If they wail and shriek and throw their hands up in the air, you'll know you're well and truly hosed...

randy

----------------
"I've suffered for my art. Now it's your turn."

Seek and ye shall probably regret it.

[www.theban.be]
[www.theban.be]

The .wav file is the 7M uncompressed version.

He's in Belgium. If we do this just right there will always be somebody working on this around the world. I got it from him. I'll pass it to Jude. She can push it to....do we know anybody in india?

Koz

Good idea about the compander in post, BTW.

Koz

Ack, that's a beauty, Koz. I'd be right off to a pro sound guy with a bucket of cash AND chocolate.

<<<I'd be right off to a pro sound guy with a bucket of cash AND chocolate.>>>

So would I. So would he. Where would you go?

Koz

Hi Koz

Here's my 1st attempt with the wav file (I'm running out of time today, so I'll play around with this some more next week).

Here's the file: www.hd-vo.com/gm2rev1.mp3

I used Adobe Audition 2.0 "capture noise profile" on a segment with no speech. I had previously changed "FFT Size" to 12000 points - that seems to work better. Then I applied that profile to the entire file under "Noise Reduction". I used the "preview" funcition to make the compromise between too little noise reduction, and too much filtering.

I then applied some "expansion/gating" to the entire file, again using the "preview" function to hear the changes as I made them.

I then mixed that file with some stock roomtone to mask the gating.

Whether your client would think this was "good enough" I don't know. There is still some room for more "noise reduction" filtering, at the expense of the voices gettting a bit "squeeky" in places.

I don't know where this audio is going to be used, but I'm pretty sure that a little background music would completely mask the remaining "artifacts".

Audition runs well on Macs with Parallels, and I would presume it would work with Bootcamp too. Next week, sometime, I attempt to do this with Soundtrack Pro.

Travis
VoiceOver Guy and Entertainment Technology Enthusiast
[www.VOTalent.com]

<<<I attempt to do this with Soundtrack Pro.>>>

You don't have anything like Audition control in STP. I was able to do much better even in the much older Cool Edit (grandpa to Audition). Yes, cranking the sample points up gives you a much better chance at success as does working on the WAV file rather than the pre-compressed 10:1 MP3.

<<<www.hd-vo.com/gm2rev1.mp3 >>>

I do like that effect. You turned it from weapons-grade garbage to a consumer grade recorder running in a slightly live room. That's way beyond anything I could do. Have you ever experimented with the technique of double reducing? Somebody posted about a way to run the noise reduction more than once and getting much better results. You must need to hold your mouth just right to do that, because it didn't help when I tried it.

I asked the poster to leave a message on here. He needs to personally put in a word or two. I'm holding hands in all directions (and time zones) and I'm going to need to go potty soon [crossing legs].

Koz

Hi Travis, hi Koz, hi everyone else,

Thanks for all the efforts you are doing to help me out. A box of Belgian chocolates indeed would be appropriate.

Quote
Travis
Whether your client would think this was "good enough" I don't know. There is still some room for more "noise reduction" filtering, at the expense of the voices gettting a bit "squeeky" in places.

I think the result is more than good enough, I'd never thought it would be possible to restore the recorded voice. Great job!

Does anyone think it is doable to explain me the steps I have to do to get the same result for the entire track (wich is about 45 mins). I don't have Audition, but I do have Soundtrack Pro. I must confess, I'm a rookie in that. (Normally the interviews I record are of good sound quality, this one just was recorded long time ago on an old cassette...

Thanks again

greets

Jan from Belgium

Hi Jan.

I was able to get similar results from Soundtrack Pro 2.01.

http://www.hd-vo.com/gm2-stp2.mp3

The process was: "Process/Convert to Mono/Left Channel" to get rid of the 2nd channel.

Then highlight a segment of the file which contains ONLY the noise you want to get rid of. It needs to be a portion without any speech, or transient noises. You want a segment containing nothing but the machinery (A/C? ) in the background.

Then, "Process/Noise Reduction/Set Noise Print" - This will capture a noise profile containing all the frequencies in the segment you selected.

Then "Process/Noise Reduction/Reduce Noise" - Press the preview button and play around with the settings... The more you play with the controls, the more you'll understand them.

When your satisfied with what you're listening to, select "Apply". I then repeated the process to get rid of some frequencies missed the first time around.

Then, "Effects/Dynamics/Noise Gate". You'll also need to play around with the settings to get what works best. In the process, you'll learn what the controls do. (It might be good to have the instruction manual handy while you're doing this.)

You might want to reduce the setttings a bit, which will retain more of the background noise, but reduce the "squeeking" you get from the voices.

More later....

Travis
VoiceOver Guy and Entertainment Technology Enthusiast
[www.VOTalent.com]