A few weeks ago, I had the chance to work on a really terrific musical. “Working” is based on a book of oral histories by Studs Terkel, and it’s a compilation of vignettes about working people — what they do, and what they think about what they do. Most of us either work or have worked, so it’s easy to identify with at least one of the people portrayed. Some of us even see a little of ourselves in several of those people.
There were eight performances over three weekends. Due to a death in the family, I actually had to eject from the whole second weekend, and was backed up by a great and very talented team who’ve worked at my side before. I felt bad about this, so when I came back for the last two shows, I brought along my computer, pro audio recording software, and 16 channels of interfaces, and recorded those two performances to multitrack.
I didn’t tell anyone (other than the tech team) that I was doing this. I didn’t want the cast worrying over being recorded, and I also didn’t want anyone to expect anything if for some reason the recordings didn’t turn out to be usable for some reason. Shortly after the show closed I did let them know, also explaining that it would take a couple of weeks before the finished product was ready. I let cast and crew hear a couple of rough mixes as I worked.
Mixing a musical live is hard and exacting work, but mixing one after the fact from live tracks is an even greater task. You hold yourself to a higher standard — theatre is live, but recordings are forever.
As I was working through the task, stealing time where and when I could, I realized it was going to take more than the couple of weeks I has initially mentioned. But I was making steady progress, the best progress I could manage while fulfilling other responsibilities.
About a week before this writing, three weeks post-show, I ran into one of the cast members in a grocery store. Meaning well, he suggested that if it was just a time thing that was keeping me from finishing, he’d gladly do some of the mixing for me. I know he bore no ill will and just wanted to speed things up, but that cut a little. In a way, offering to do my work made me feel just a little less like my particular talent was seen as uniquely valuable.
Yes, time is a factor, and it takes a LOT of time. But equally important is the experience, skill, and knowledge that comes from a lifetime of mixing. And I’m a bit of a perfectionist.
Just the opening number, which is four and a half minutes of audio, required about two and a half hours of work just to achieve a presentable rough mix. Getting to a final mix I was proud of took two more hours. Four and a half hours to mix four and a half minutes of audio might sound outrageous, but one must remember that in the music industry, it’s not unheard of for an engineer and producer, working together, to spend twice that amount of time mixing a single song.
Why does it take so long? Well, in the spirit of “Working,” let me talk about my job. Here’s how I go from raw tracks to a final mix, my workflow.
First, I have to audition the tracks from both nights. I’m listening for general quality, dropouts and mic issues, and issues with the performance itself. I’ve got to choose which of the two nights represents the best performance and the fewest technical hurdles to overcome — for each song individually. When I’ve made that choice, I then cut out the tracks from that song and that night and move them to a separate project file to work on. Backups are made immediately.
The second step is going through the tracks of the song, second by second and line by line, doing any necessary normalizing or gain changes, and listening critically to each vocal. If there’s a dropout, a mumbled line, a missed line, or a wrong word or phrase, my go-to is the other night’s recording. I cut out the offending piece and splice in the same line from the other performance. Making these cuts as unobtrusive as possible is part of my art. Sometimes they don’t match well; the mic was in a different position each night or the tempo was ever so slightly different. I’ve spent upwards of 45 minutes replacing one line and making the edit seamless and inaudible, automating EQ changes and time-stretching or time-compressing the clip, and fine-tuning the level match with fader moves or region gain changes.
My third step is checking pitch. I don’t believe in pitch correction (auto-tune). My threshold for feeling the need to use it is really high. But I want these recordings to be something the cast can be proud of, so if someone goes really badly off on a note or two, I’m willing to fix it. I will not auto-tune a whole track. So my pitch corrections are done manually, a note at a time, and carefully manipulated so that they’re not TOO perfect. If an actor went way flat on a D, I look back at the last good D he sang. If that one is only a few cents flat, I match that on the one I’m correcting. I do this while listening to the phrase loop by, until I’m happy with the notes in question.
My fourth step is scoping down the tracks. Theatres are full of reflections. If all the mics were open all the time, all sorts of undesirable ambience can creep in. So my next task is to go in and cut out sections of each track that are not part of the performance. If someone isn’t singing this particular song, I delete that entire track. This reduces the amount of work I need to do in the next step. In areas where actors say or sing a line at a time, I cut out the unused portions of the tracks. This doesn’t really reduce the number of automation moves I have to write, but it does give me a visual reminder of where the lines are.
In the fifth step, I apply EQ and dynamics to each vocal and instrument. These become templates, but sometimes must change between numbers. For example, when a person’s role changes from ensemble to lead, as happens often in this show, the voice will need different dynamics. If you’re an actor, you can think of this as an extended mic check, with much more stringent standards. It can take a while to dial each voice in perfectly.
Finally, I’m actually ready to mix. I make a few passes with all the vocals muted, just to get the band where I want them. The keyboardist changes patches now and then and I have to compensate. Sometimes the guitarist hits a chord, realizes he’s too loud, and turns down a little; I have to match that in reverse to get steady levels. The drums, which sounded great acoustically in the house, were miked only for the recording, so the track isn’t perfect and has to be processed to remove extraneous sounds.
Once the band’s locked in, with moves written and enough headroom left for the vocals, it’s time to bring in the voices. The fader moves I normally make during the live show are reproduced as automated fader moves in my computer’s mixing software. I can write them on a timeline by placing influence points, or I can move virtual faders, with the computer memorizing and reproducing them. I can trim a move by dragging a line or by touching a fader. It’s good that there are so many ways to accomplish the fine moves required. It can get incredibly tweaky, making lead vocals clearly heard and background vocals all match up. The gain changes I made earlier have reduced the number of moves I have to make, but it’s still complex.
Next I add post-processing to the final mix, which is normally nothing more than a multiband compressor that’s been set to even out the rough edges (I don’t leave many) and tighten up the level so all the tracks are consistent and volume-matched. It’s what in my workflow corresponds to mastering. Then with the click of a button I print the mix to both an uncompressed WAV file and a 192k MP3.
Off I go to my living room stereo, where I listen to the mix on that system. Then I take it out to my car and listen there. I listen to one pass on headphones, too. Finally, I play it out through a pair of cheap-ass old Pyramid speakers that sound like crap. I want to be sure the mix sounds as good as possible in each of those places. If it does, I’m done and can move to the next song. If not, I reopen the mix and make appropriate changes (a little less bass here, a little trim to the vocal EQ to fix a little tinniness), and print new files. Lather, rinse, repeat.
And of course, I’m still not completely done. The next day I usually replay the mixes again. This is an idiot-check, just to make sure I didn’t miss something because I’d heard the song 247 times in a row. “Wow, the keyboard really came flying out there, didn’t it? How’d I miss that yesterday?” Another tweak, another print. And then I’m done — but only with that one song. Then it’s on to the next number.
It takes time, energy, and patience, but I love what I do. I came home Sunday afternoon after mixing a church service and worked for four and a half hours on this before my ears were shot. I stopped halfway through a song, and that was a mistake. I heard that song, and the parts of it that still needed work, in my head all night. I slept very restlessly.
On Monday, a holiday, I started at about 1 PM and mixed until 10 PM with only a short break for dinner. I went to bed happy to have finished complete final mixes for all of Act 1. It was a good time to give myself an intermission. 9 hours is a long time to work, and ear fatigue is a very real thing. I was mentally exhausted, too.
Evenings this week will doubtless be spent completing Act 2. I expect there’s another 20-25 hours of work to be done there. I’m sure someone with skills and talent could help, but I really want the whole show to have a consistent sound, the way I hear it in my head. And it means a lot to me to be doing this. I want to see it through. So if you’re one of those cast members waiting to hear this and you’re wondering why it’s taking so long, now you know. Be patient. I will probably let you hear Act 1 this week.