Recently I was hired to finesse a rather rough live recording of Michael Franti‘s performance at the protest concert outside the Republican National Convention in 2008. The recording was done with a pair of shotgun mics placed fairly close to each other.
In fact, the recording was essentially mono – probably because of the mic positioning and because the live mix was essentially mono. The recording was almost entirely the output from the stage mains, which managed to be both heavily compressed AND had tons of volume fluctuations.
I heard many problems with the recording. First of all, it was extremely cloudy due to a heavy build-up of low mids. Second of all, the volume fluctuated throughout the performance. It also suffered from compression induced volume inversion, whereby the mix was so compressed that the loud parts of the song were quieter than the soft parts of the song. The guitar had almost no body to it, mostly consisting of the pick attack in the 3.5k-6k range. The vocals, thankfully, sounded pretty good.
Finally, all the recording’s flaws were amplified 3x by the fact that the recording was essentially mono. Our brains aren’t used to processing mono material (especially these days) so it made things sound a billion times worse.
First I EQ’d the mix with the Liquid Mix Massive Passive model, chopping out a good 6db of 300hz with a very wide Q. Then, I rolled off 25hz and below with a high-pass filter. There was a subtle subsonic rhythmic thud that I heard, maybe Michael stamping his feet on the stage being transmitted through the mic, that I wanted to maintain. But, there were some low low subs taking up too much energy in the mix that would bother my compression choices later. Then I boosted 10k+ a tiny bit – like 1.5db. The mics sounded like dynamics that didn’t go much higher than that.
Next I threw up a Waves de-esser to take the edge off the jangly guitars. I dialed that in at 5k, set it to high-band reduction, and had it suck out 4db when the threshold was hit.
My buddy who was originally trying to work with the audio tried lots of things to improve the dynamics, including messing with multiband compression and expanders. I’m not a fan of those sorts of things, but I decided that the dynamics did indeed need to be addressed. It would have to be subtle and gentle, as the material was already smashed.
Firstly I pulled up my Liquid Mix model of a Summit Audio TLA-100. I like this compressor for almost everything because it’s warm and extremely smooth. But it wasn’t smooth enough for this application – around 1.5db of slow compression. It made things worse.
So I went even further and pulled up the ultimate ‘natural’ compressor model – the Fairchild or ‘primitive’ model in the Liquid Mix. This did the trick – warming the track a bit, controlling the dynamics, and taking the edge off the guitars and vocals.
Now the toughest part: dealing with the monophonic sound. I could try to spread the recording out with a stereo imaging plugin, but I don’t like to go too far with those as the sound starts to thin and get phasey. I decided stereo imaging wasn’t a good choice.
Since this was a live recording I had more flexibility. I decided to approach it like a post-production mix for a film. The outdoor performance had a good amount of ambience to it, so I took it a step further. I pulled up a ‘scoring stage’ reverb impulse in my Waves IR-1 and brought the pre-delay up to 100ms to give it more bounce. Then, I threw up a stereo delay, one side about 20ms apart from the other, and blended that in just a smidge panned about 80% to simulate slap-back from buildings and walls. This technique only worked because of the extremely high quality realism of the IR-1 model. For shiggles I tried other artificial verbs and IR-1 models and none of them sounded anywhere near as genuine. A Lexicon-esque verb was way too smooth and didn’t have the early reflections needed to place the sound. I got lucky on my first choice.
Suddenly the track sounded like it was happening live, not just a documented approximation of the performance. The verb sounded really cool during the portions where the audience clapped and sang along. The applause at the end sounded bigger thanks to the delays.
A/Bing with the original was huge. They sounded like totally different recordings. It still may not be good enough to pass QA at Sony but it’s about as good as it’ll get. And it’ll ‘sound’ better when put to the footage they have.