Audacity effects for cleaning up a Skype call podcast?

April 23, 2008 9:29 PM

I've recorded a Skype call via Audio Hijack, and my plan is to podcast it. Now I'm editing in Audacity. What effects/compression settings should I use to improve sound quality?

The source file doesn't sound too bad, really, but I'd really like to see if I can improve this -- at least to remove mild hiss and improve some vocal fidelity.

I've dug around online and haven't seen anything that isn't meant for semi-serious sound designers...which I'm not. Maybe that's the problem. Suggestions?

posted by diastematic to computers & internet (12 comments total)

If you've got hiss, you don't want a compressor (which would bring up the volume of the hiss, compared to the louder parts of the signal), you want the opposite - an expander (or noise-gate). Used delicately, it could clear our your hiss... but used too heavily and you'll start chopping up the talking.

I'd start by normalising, then start fiddling with a noise gate.
posted by pompomtom at 9:43 PM on April 23, 2008

Throw a 4 band EQ on there, take the third band (probably set to default around 10 Hz), narrow the Q to zero, and crank the gain all the way up. Now sweep the frequency up and down until you hit the spot where the hiss is the loudest, and drop the gain about 6 or so db. If you need to use more bands to rid of other hot spots, do so.

The human vocal range (while talking) is about 80 Hz to 1100 Hz, so if you want your voice to be clearer, find the frequency in which it lay and boost it (just a tad) with a fairly wide Q, this will make it sound a bit closer.

As for compression, you'd really only need a slight squeeze- if any. I'm guessing the dynamic range of the call isn't so drastic that you'd need the quieter chatter much louder (compression makes things louder, not louder things softer, remember- sometimes to detrimental Britney Spears-sounding dimensions), so try just boosting the overall gain and if that doesn't work- then I'd throw a compressor on there with a quick attack/release, low threshold, and a low compression ratio.

Cheers
posted by pedmands at 9:52 PM on April 23, 2008 [1 favorite has favorites]

And the slight compressor would minimalize the boost of the remaining hiss, as pompomtom wisely pointed out.

Audio can be attacked from many angles, it all depends what you want the end result to be exactly. No hiss at all, or level speaking (i.e is the person on the other end too quiet to hear)?
posted by pedmands at 9:55 PM on April 23, 2008

The bandwidth of a telephone line is 8kHz, so you can start by EQ'ing away everything above that, then play with an expander to get rid of the interphrase noise. It might sound better if you roll the bass off a bit if it winds up being muddy after both the previous steps.
posted by rhizome at 9:56 PM on April 23, 2008

If the people on the recording show up at different volume levels (pretty common in a telephone recording), consider using Levelator. It was created to adjust everybody to the same level. It's a stand-alone program, operates on WAV or AIFF audio files, cross-platform, and free.
posted by exphysicist345 at 10:17 PM on April 23, 2008

Slightly off-topic but related: you might want to check out iZotope's excellent guide on mastering. While you are not doing mastering, the PDF does cover topics on EQ and compression that would be of interest to you, since these two are essential to the mastering process.
posted by seeminglee at 11:44 PM on April 23, 2008 [2 favorites has favorites]

The next time you need to record a conversation in this manner, you might want to take a look at mathowie's "how to record a kickass podcast between two macs and cheap". It requires more setup work and tech knowledge on both ends of the call so it's more suited to a "co-host" situation than an interview but maybe you'll find something useful.
posted by bcwinters at 5:09 AM on April 24, 2008

run it through the Levelator ? Supposed to work wonders on voice recordings automagically.
posted by namewithoutwords at 5:48 AM on April 24, 2008

Also, to remove the hiss, you may want to try to get a sample of the hiss on its own (no noises on top of it) then reverse the polarity of that and run it on a second track. That way, the original hiss and the reversed hiss cancel each other out.

However, depending on the frequency range of the hiss, that may end up being detrimental to the voices recorded. Looks like you've got some experimenting to do!

I would second the EQ idea for early work and then fiddle with any other ideas you have to get to an end result you are happy with.

After further googling, apparently the Noise Reduction tool in Audacity is quite good. i haven't tried this myself but, here are some instructions on how to remove hiss inside Audacity
posted by moocheen at 6:11 AM on April 24, 2008

Although, I would take the advice in the last paragraph - especially

"if you are unable to remove the noise completely without distorting the voice, you have a problem with your hardware and need to have it checked by a competent hardware guy"

with a grain of salt. This is just flat out untrue, you aren't going to get a perfect result.

It's almost as good as the final sentence

I've delivered studio-quality sound to clients using nothing more than my home computer, a $2 microphone and Audacity.

I didn't know you could buy a microphone for $2!
posted by moocheen at 6:20 AM on April 24, 2008

(compression makes things louder, not louder things softer, remember- sometimes to detrimental Britney Spears-sounding dimensions)

I'm not sure if this is what you meant to say, but compressors do make loud things softer. One usually boosts the overall gain after compression for an increase in average level, but a compressor works by squashing the loud bits to be more even with the quiet bits.
posted by ludwig_van at 9:42 AM on April 24, 2008

A few extra thoughts:
It's VoIP. That means some fairly heavy voice-specific codecs, which are going to do all sorts of odd things re: hiss, voice, & sidetone. Specifically, they'll tend to gate background noise heavily yet add their own hiss (to make it sound "normal" again), have non-linear attack & decay profiles to minimise the required bitrate without affecting intelligibility, and generally "make up" a lot of the stuff which makes a phone call sound like a phone call.

All this makes it very hard to clean it up easily, if at all. Personally, I wouldn't do any more than maybe add only the very slightest bit of stereo reverb to give it some presence and, depending on how it fits into the rest of the podcast recording, a touch of echo / ringing to accentuate the fact it's a phone call.

With something that damaged, you're getting away from "how do I clean this up?" and into "how do I subtly bend psych-acoustic effects to my advantage?" territory...
posted by Pinback at 5:17 AM on April 25, 2008