Noise reduction

I’m looking for some tips or ideas to help with noise reduction. I’m using the latest Audacity Beta. I have a voice recording captured with Camtasia and i’m working with an exported .wav file. Most of the noise is computer/fan noise. I’ve selected an area with no voice to use as a profile, tried different settings. In the areas where there is no voice, the noise is reduced pretty well, but when the voice starts, the noise level increases.

Any advice on how to better remove this noise? I also need to boost the levels as the recording is a little quiet. The audio track is Mono 22.050 kHz


<<<I’m using the latest Audacity Beta.>>>

Hard numbers are good. I use 1.3.4. Which one do you have?

If you had one problem you’d be OK, two problems is a little rougher, but you pretty much did everything wrong.

You have a low level mono recording with a poor sample rate and you have significant interference from whining computer fan noise.

Do the sample profile again and this time apply it with 18dB reduction, 200 to 500Hz frequency smoothing and no delay. If the show gets a little better, increase the reduction until it starts pumping like you had. Mess with the frequency smoothing down until the show starts gargling.

If you never get anything usable no matter what, you get to shoot it again, and this time move the computer out of the room, get closer to the mic and choose a much higher bit rate. I use 48000/16-bit (the video standard) for live capture. You can always sample down, you can’t sample up, and MP3 and other encoders work far better when you give them a perfect, high quality show.


Thanks for the advice, those settings helped a lot. Unfortunately i can’t move the computer, well not by much anyway. The settings for the audio are specific settings i have to use for recording video tutorials for a 3rd party. I’ll have to get creative next time to try and reduce the noise before i record.

Thanks again.

You can play tricks with furniture pads, pillows and hassocks. You can’t block the computer air flow, but you can record behind a pile of throw rugs and that will help. I have done voice tracks in my carpeted bedroom with the microphone set up on the crazy quilt on the bed. That’s how I recorded the Left-Right track from here…

There’s no processing (other than the last track intentionally damaged). That’s a straight recording.

If you’re using an omnidirectional microphone, you might experiment with a directional unit and point the “dead” side of the microphone in the direction of the worst noise.

You can also experiment with headset microphones. If it’s good enough for Britney Spears…

You can record in very high quality, process and edit and then downsample for delivery to the client. The tools all work really well on a high quality show.


Thanks Koz, a lot of great advice there.

If the fan noise is at a constant frequency you could try a notch filter.
The “computer noises” (clock noise/sampling artifacts) they too can be notched out.
Use “analyse” to plot a frequecy spectrum of a “no voice” bit, if you see a big peak on the graph try notching it out.
A notch filter plug-in for Audacity can downloaded here…

BTW you can maximize the volume by using the “amplify” effect to boost the levels.

Thanks Trebor, i’ll give that a try.