Mark Willsher on Acoustica 8 - Keep the Best Take, Lose the Noise

Mark Willsher has spent his career in rooms full of people playing at once. The Grammy-winning engineer and mixer's credits run from the San Francisco Symphony to the scores for Ultraman: Rising and The Lord of the Rings: The War of the Rohirrim, and on any given project he might be setting up the mics, mixing a session someone else recorded, producing, editing or mastering. Often it's a combination of those.

With fifty or more musicians in a room, perfect silence is a fantasy. So for 25 years Mark has leant on spectral editing to keep the take that sounds best instead of the one that happened to be quietest, and these days that work happens in Acoustica.

At what point do you get involved in a project?

"It entirely depends on the project. I tend to wear many hats, so I can be on extremely early, or I can be on a couple of weeks before recording. It varies depending on the project and also depending on where we record."

Some rooms are more unusual than others. On one project with Emmy-winning composer and collaborator Stephen Gallagher, Mark recorded in the chapel of a small convent in New Zealand, with the choir singing towards the microphones and then turning to sing away, before morphing between the two sounds. 

What makes orchestral recording different from a band session when it comes to noise?

"Even if you're doing strings on their own, you've got maybe 50 players in the room, and they're all together. So it's not like you have complete control over every individual instrument. You've got a lot more opportunity for extraneous noises, and for click bleed if you're not careful when you've got 50-plus people with headphones on, sometimes those headphones may bleed. And in a lot of the rooms we record ensembles in, you hear outside noises, an ambulance goes by, someone blasts a horn, you hear the London Underground Jubilee Line below."

When did spectral editing come into it?

"Back in maybe 2000, 2001, I started using Algorithmix reNOVAtor, which was one of the first spectral editing tools. What was instantly amazing was that I didn't have to worry about a lot of those noises. Rather than living with a phrase that wasn't the best performance but was the quietest, I could use the one that was the best and just remove the fact that someone squeaked their chair, or a horn player blew out their horn, or some other unintended noise that becomes distracting."

That must change things for the players too.

"Exactly. It's terrible when you have something that was just exquisite, there was a noise, and you need to do it again. You end up doing it three or four more times because they never get back to what they had. Then you have to move on because of time, and no one's satisfied. The players know they didn't get there, and you know you didn't get there. Knowing you can fix it after the event is definitely calming."

The end of a take brings its own problems. "Ideally you want at least a second after the decay ends before there's any noise. But in a room with a long decay time, the moment there's no sound happening anymore, there'll be rustling of music and moving of chairs. No matter how many times you say it. There always ends up being some cleanup at the end of things."

What are the most common problems you deal with in post?

"With strings, someone accidentally hits a wrong string. They don't play a note, but their finger touches it and you get a resonance of a pitch that shouldn't be in the chord. Or players check their pitch very lightly before a section comes in, and of course the microphones pick up everything. Then of course there are page turns. We do everything we can to minimise them, but a large group of players turning pages at the same time makes noise. And a lot of American orchestras record from live concerts, because studio sessions are too expensive, so you get audience coughs and people unwrapping sweets."

Sustained noise is rare in the controlled spaces Mark works in. The hours go on transients, like something dropped or a chair dragged across the floor.

How do you remove a noise that crosses a sustained note?

"If you've got something sustaining with a bunch of harmonics and there's a noise crossing them, what's worked best for me is to go in between those harmonics, select the clear space where there isn't any noise, and copy that over the noise without touching the actual note. Then do that for every single one. You can get it very transparent, but it's very time-consuming. If you accidentally process even just a harmonic, it's blisteringly obvious that you've done something."

Why did you move on from reNOVAtor?

Mark used it for close to 20 years, but it only worked the way he wanted inside Pyramix. As more of his work moved into Pro Tools, the only reason left to open Pyramix was cleanup. Then six or seven years ago he forced himself to try iZotope RX.

"In Pyramix I could load a 60-microphone orchestral multitrack into the spectral editor and work across all of those tracks while listening to my mix as it is on the mixer. There was never any guesswork. I thought I couldn't work without that. What I discovered was that it wasn't as important as I thought, as long as I could process all the tracks as a unit."

He explains his reason for treating noise at the individual track level. "I have colleagues who print the mix and then process the mix because it's easier. The problem is that if you get notes and change the mix, you've got to do that work again. I don't like to do work twice. Anything that happens, I want to do in the multitrack, so moving on, it's always there." 

How does that look in Acoustica 8?

"I'll take ten tracks at a time, say the main array, send them over and work across all ten at once. If I can still hear the noise, I'll figure out which other mics it's on and take those over. If a wind player hits their mic stand, I'll clean that mic and the main array, and the noise isn't really audible in the other spots."

Batching ten at a time sounds slower on paper. Mark says it isn't, because the work itself goes faster.

"I can get away with less fussy, precise work and do more broad, overall stuff." He is confident that Acoustica is the tool to support this workflow. "If I had to guess, I'd say the interpolation is better in Acoustica, and it probably has a better understanding of what is background noise and what is content. I use the feathering tool a bit, but I find I don't really need it all the time, it's so good, the feathering tool is a bonus."

Anything smaller that makes a difference in Acoustica 8?

"Something that seems simple, but I haven't had before: when I select something and hit play, the default is that I only hear what I've selected. That makes it much faster. When something's complicated on a spectrogram, you're asking, is that the problem or is this the problem? With this I get there faster, both because of that instant solo and because of the processing itself."

Most of his time is spent by hand in the spectral editor, though he reaches for De-click and the improved noise reduction in DeNoise 3. "I find its automatic detection is pretty good. I don't find myself fussing with the settings as much. It identifies the issues, fixes it, sounds good, and I move on."

His verdict, after years with other tools, is blunt. "For a long time I've used the iZotope stuff, and all of these Acoustica modules are so much better." 

You're not a typical user, though.

“Probably not. More of my time is manually in the spectrogram rather than using the various tools. There are a few workflow things I'd like to discuss in terms of speed in the spectrogram. But because of the instant solo and how good the algorithms are, some of those things I don't care about, because I still get the result faster.”

Thanks so much to Mark for sharing his insights with us. You can find out more about Acoustica 8 below.