Noise coming from DSLR objective's Optical Stabilization

[diffler]

I recently filmed an interview and started editing it. When listening to the audio, I quickly noticed I had screwed up. I had left the OS turned on on my Sigma objective and the motor kept buzzing in the background through the whole interview. I can't redo the interview.

Can the clip be saved? Here's an example of the background noise.

https://drive.google.com/file/d/0B_NuJGc2xp8tUFotSkdobktPYzA/edit?usp=sharing

I'm working on Premiere and tried few methods I usually remove background noise but this is different frequency and I just can't do it!

Thanks a lot!

[smrpix]

This is tricky because you also left the AGC on, so different noise frequencies are emphasized even within your short sample. Automatic noise removal methods are only partially effective at best.

You're going to need some combination of audio gates and noise reduction to get close, and you're probably going to have to do it in very small pieces.

The real solution going forward is to never use a built-in or camera-mounted mic for interviews.

[jagabo]

Use a narrow notch filter at 670 Hz and 2000 Hz. Another at 1340 might help too.



Those are two notches using the equalization filter in Audacity. You can probably more accurately determine the exact frequencies and get even better results using the notch filter.

[Cornucopia]

I thought the 1st fundamental was more around 630Hz.

This one I did in Audition. Process went as follows:
1. NR Profile of noise alone
2. Process Broadband NR using profile as guide, ~50% mix & ~-30dB cut (IOW, kind of mild).
3. Notch filter at 630, ~-25dB, Q=~25 (but with actual rise slightly lower & higher to perceptually compensate for the loss of vox in that range)
4. Non-linear dynamics (S-curve compand), mild
5. 2nd Profile but restricted to higher harmonic (~2200?), NR applied to it.

Here's my sample...

BTW, You SERIOUSLY need to NOT use a camera mike. Interviews are best done either Boomed Shotgun (1st choice) or Lav (2nd choice). Both options are better wired than wireless. Location (closeness) makes all the difference. Here is a demonstrating example of an interview I recorded just over a 1 1/2 years ago, using a Sony Wired Lav + Shure FP33 mixer + Zoom H4N...

Scott

<edit>I think I like jagabo's a little better than mine! I hate how FFT-analysis-based NR "garbles" the sound a tiny bit, even when used mildly. I was just trying to get a little more aggressive with it, so that's why I went with it. jagabo seems to have left a little hiss in, which retains some "air"</edit>

[diffler]

Thanks for your help everyone.

First...woah! Those sound very good jagabo and Cornucopia! I'm amazed how well you took out the disturbing noise Cornucopia. But indeed, I'm looking for a bit busy feeling and for that your clip is perfect jagabo. Thanks a lot you both, with this it's publishable, I'm very grateful!

And I know I shouldn't use the camera mic. I thought I'd leave the answer here, but woah Cornucopia! That sounds VERY good! I'm very new to video filming and processing and I've somewhat ignored audio, it was enough if for example the voice of a person is audible. But if I compare your clip to mine....woah. I'm embarassed. Even if I don't have to get pro quality, I can't go on like this. I'll try to muster up some finances and get a real mic.

*bows deep* Thanks again! You saved me!