For those who know anything about audio…

I’m working on encoding videos from the big Ship of Fools / Andy Ober Orchestra show this past March. The video’s coming through OK (though I’m still fiddling with mencoder’s various knobs to get a good-looking video without taking up too much disk space).

However, making the audio come out well is giving me some problems. I fear there’s no real solution to this, but I’m going to throw this out there in case someone reading this in the near future knows anything about this topic.

Here’s the situation: I have a live recording that has four basic sources of sound in it: a keyboard, an electric guitar, vocals and speech (primarily but not exclusively male), and an audience. On the recording, the instruments and audience come through quite well, but the vocals and speech are way too quiet, sometimes to the extent that they’re completely drowned out. (The imbalance was a bit of a problem live, but on the recording it’s much worse.)

I’d like to somehow remedy that, or at least ameliorate the problem, so that you can actually hear what’s being sung or said. Unfortunately, everything’s on one audio track, which means I have no simple way of adjusting a single source. Whatever I do, it has to be done on the entire audio stream.

So, is there any way to make the vocals and speech louder without distorting everything else too much? The only thing I can think of is to tweak the audio equalization to make some frequency ranges louder and others quieter. Is there a frequency range that primarily only the vocals will fall within? A little random experimentation suggests not, but this is definitely out of my area of expertise.

Or might there be some other trick I’m missing?

Of course the “proper” solution would require having recorded separate audio tracks for each instrument and vocalist, but that presupposes access to equipment we don’t (and didn’t) have access to. And it’s too late to argue about that now, since the recording’s already been made.

The recording is mostly servicable as it is, but it’s hard to fully appreciate song parodies when you can barely hear the lyrics.

I promise a free copy of the video to anyone who offers a good solution. (Pay no heed to that it’s my intention to distribute the video free to the world once I’ve got a good encoding.)

7 Responses

  1. While you would have to use proprietary software on a Windows machine, Adobe Audition (formerly Cool Edit Pro, I believe) comes with a number of pre-set filters you can run on audio tracks (or portions thereof) to enhance / dampen common frequencies. I was able to use it more or less out of the box to get a singing-free version of Tom Jones’ “Sex Bomb” for karaoke purposes.

  2. For karaoke purposes aye? For some reason I feel like you’ve used the alyrical version of “Sex Bomb” for purposes other than karaoke purposes.

  3. Mencoder, I believe, lets you adjust the levels of different frequencies, but it has to be specified on the command line and there aren’t any presets. I don’t suppose you know what frequency ranges I should be targetting, do you?

    I wonder what kind of video editing software they have on the PUCC ITaP machines….

  4. I’m not sure they would make that information too publicly available, since it is a commercial product and all. But if ITaP doesn’t have it, and google searches on removing the vocals from songs (which is fairly common is will be the opposite of what you need to do) turn up nothing, then I could take a whack at doing it myself in Audition over the weekend.

    And Benji, it’s true. My remix of Sex Bomb going to be my “walk-in” song when I start Pro Wrestling later this year as “The Bolshevik Balladeer.”

  5. According to Google, indexer of all human knowledge, vocal eliminator tools work under the assumption that you have a stereo studio recording where the vocals are essentially identical on both channels but the instruments aren’t; by inverting one channel and mixing them together into a single mono channel, you cancel out anything common between the two. (for further reading, see this or this)

    Alas, that technique doesn’t work on this recording. Although it’s in stereo, the vocals aren’t the same in both channels, so nothing really gets cancelled out. (Or, if it does, it’s really subtle.) Any analogous trick to accentuate the vocals isn’t likely to work either.

    However, if you do this vocal elimination trick on the “studio” version of Pi Pi Pi, you end up eliminating the guitar and the main vocals. I call the result “Pi Pi Pi (piano versus the muted echo chamber mix)”.

  6. I still do not understand why you can not just boost the sounds as seperate tracks. The you can simply mix them down into one track. So you can destroy all other sounds if need be in one version of the audio, and then still maintain them in another copy.

  7. They aren’t separate tracks for each sound. There’s just a left channel and a right channel. That’s why this is perhaps an insoluble problem, rather than something that can be fixed easily in an audio editor.

Comments are closed.