When Doug Kaye announced the availability of The Levelator, I hoped it would spare me the worst drudgery of podcast production. After recording and postprocessing my next Friday podcast this morning, I'm convinced that it will.

Here's the deal, from my perspective as an audio newbie now plunged into the deep end. As my podcasting method has evolved, I've settled into two modes of editing. In one mode, I refine the content of the recording. That involves fine-grained internal editing -- trimming out excessive ums, uhs, and pauses -- as well as coarse-grained edits that remove less interesting passages in order to focus on the most essential parts of the conversation. Applying both methods typically reduces the final product to somewhere between 70% and 90% of the original length and, in my opinion, sharpens the result in a way that's well worth the investment of time. I've always enjoyed this kind of editing in the textual realm, and it turns out that I enjoy it in the audio realm as well.

The other mode involves the purely technical work of taming sometimes-noisy phone lines and evening out audio levels. As I've become more sensitive to audio quality, I've found myself spending more and more time on the leveling process. It's not only needed to balance the caller and the callee. There can be a ton of loudness variation just within the caller's track. When you start fiddling with that, you're on a slippery slope that leads straight into a pit of drudgery.

In a talk given at the Podcast Academy, Daniel Steinberg noted that the normalization function in audio editors like Audacity and Audition "doesn't do what you think it does." I guess I'm not the only one who assumed, incorrectly, that normalization's job was even out loudness variation. But apparently it doesn't. It levels off peaks, but doesn't raise up valleys.

Levelator does do what many people expect normalization to do. It brings the valleys up close to the peaks. It's going to spare me a ton of the kind of editing drudgery that I'd outsource to an assistant if I could. And it's going to enable me to focus on doing the kind of editing that I wouldn't outsource even if I could. (Well, I'd outsource the fine-grained internal editing, but not the coarse-grained editing.)

The credit for this excellent hack goes to GigaVox engineers Bruce and Malcolm Sharpe. Here's part of the backstory from Doug's blog:

Bruce, with help from his son, Malcolm, had proven that he knew how to tackle these problems in ways that no one else anywhere in the audio/software industry has done to date. So I asked him, "Bruce, do you you think you can write a leveler that corrects for medium-term variations in loudness instead of the short-term and long-term variatons processed by compressor/limiters and normalizers, respectively?" Bruce and Malcolm took on the challenge, and eight months later we began testing The Levelator.

I've been a professional audio engineer longer than I've been in the computer industry -- that's a long time -- and believe me, there's nothing else like this out there. I guarantee it will blow you away or double your money back. (Oh wait, it's free. I forgot.) We previewed it for 100 people at Podcast Academy 4 today, and they were unanimously impressed. You will be, too.

Indeed I am. If, as Doug says, nothing like this has been done before, then why not? My guess is that it's one of those innovations that can only come from folks who straddle two worlds. There aren't many folks who can comfortably straddle software and audio, just as there aren't many who can comfortably straddle data and voice networking. That kind of dual citizenship is incredibly valuable.

Update: It's a small world indeed. Turns out that Bruce Sharpe was CTO at SoftQuad and now heads up the XMetal group at Blast Radius. Podcasting tools are just his hobby, he says. Here's his bio from the XML 2002 conference.

Former URL: http://weblog.infoworld.com/udell/2006/10/12.html#a1543