Piano Recording Method
Mixing & Mastering
Paul's new album is out! The Broken Mirror of Memory, winner of the International Clarinet Association composition prize. “Profound and deeply moving” —Eric Mandat in _The Clarinet._ Now available.

Mixing and mastering, the refining of a recording's sound after the recording is already made, is a whole art in itself. “Mixing” refers to the combination of instruments into a single combined audio file; “mastering” refers to fine-tuning the sound so that it works on different speakers (as well as a host of other tasks related to releasing it in various media). When recording a single instrument, the distinction between these two tasks is a bit blurry. I can only begin to give you a general idea of what I know in this web page — and what I know barely scratches the surface of the field.

I've said it before in this guide, but it bears repeating: every piano, every space, every recording setup, every piece and performer is different. My strongest advice is to experiment, learn, and explore for yourself. This guide only gives a sense of things you might try, not definite answers that will work for you immediately.

The best thing of all, if you can afford it, is to get somebody who knows what they're doing to do your mastering. This probably isn't an option if you're regularly producing recordings for a podcast, but if you're cutting a CD, it's a good idea. Why? Because the a mastering expert almost certainly has better equipment, better knowledge, and better ears than you (or me)!

That said, if you're trying it on your own, the mastering you do on a piano recording is typically going to deal with these four areas:

This guide does not explain the details any of the basic concepts; I assume you can search the web to learn more about EQ. What I describe here is the particular way I apply audio tools to my piano recordings.

This is the processing chain I use for my recordings:

L/R EQ with Firium

L/R Equalization

Some of you have come to this page because you typed "eq for piano recording" into a search engine. I have some bad news for you, and I'll just get it out of the way right off the bat: Contrary to what your little boom box with the cheesy blinking lights thinks, there is no generic formula for EQ. The purpose of EQ is to correct discrepancies between the sound you have and the sound you want ... so, your EQ settings will depend on what you have and what you want, and only you know that. Your piano in your room playing your music recorded with your mics is a completely unique problem, different from every other recording in the world. There is no "piano" setting that will make it suddenly sound good. EQ is a hard problem, and nothing will magically solve it for you. Sorry.

Many of the words we might use in describing why a recording doesn't sound quite right — murky, muddy, muffled, bright, tinny, crisp, harsh, nasal, boxy — describe EQ problems. Part of my goal in EQ is to avoid all these problem qualities, and make the recording sound like the piano sounds in real life — or in my imagination.

There's a second goal: because I use different mics in very different locations, I get two very different sounds from them. I start with the M40 on the left and the GT66 on the right, and the other goal of EQ is to make the left and right sound more like each other — when they do, they fuse better into a coherent stereo image. So I apply different EQ to the two channels.

For doing mastering EQ, I use an excellent but apparently now defunct plug-in called “Firium.” The EQ algorithm it uses preserves the crispness of the transients in piano recordings, and it has a really elegant user interface (a rarity in the audio world). It's one of the few pieces of pro audio software I enthusiastically recommend: get it if you can find it. It's excellent.

The software is easy to describe; the technique is not. I can't tell you what your recording will need, but here are a few ideas that might help you in your EQ quest:

Every little bump and curve in the screenshot up above is the result of these sorts of techniques. Keep in mind that your own EQ will probably look completely different!

Stereo Image Munger

From L/R space to M/S

Our ears don't just hear left and right separately; we actually perceive the interactions of the two as an integral part of the sound. (For some fine examples of this, get some of Diana Deutsch's musical illusion CDs. They are fascinating!)

One way of getting at this interaction is transforming the sound from left-right (L/R) to mid-side (M/S) space, making adjustments in M/S, and then transforming back again. I wrote a little Audio Unit for doing this, Stereo Image Munger. I drop a SIM in the change, and have it put the mid on channel 1 and side on channel 2; another SIM later in the chain will convert back.

My first step in this world is to boost the mid channel just a bit. This helps further fuse the two very different microphones into a single unified sound — instead of sounding like two dissociated sounds coming from the left and the right, the sounds appears to come from many directions.

M/S EQ with Firium

M/S Equalization

That whole discussion above about L/R EQ above applies to M/S. Yes, the possible tweaking is endless. Yes, you will go insane. It is OK. Embrace the insanity.

I make two completely separate EQ adjustments in M/S space, which I actually keep in separate Firium instances (at left). The first is a correction for a big murky "woof" I get with my particular setup.

M/S EQ with Firium

The second is perhaps more generally applicable, a trick from Greg: I boost the very high frequencies in just the S channel. This gives the sense of a more open space — but once in a while, it can make a strong attack sound too harsh. So I use Firium's states to create a continuum of more and less of this boost, and automate the state parameter, reducing this effect in those harsh spots.

Stereo Image Munger

Back to L/R

It's been fun, but it's time to bid M/S space goodbye and head back to L/R. (In certain circumstances, when the left and right are so wildly different there's just no way they'll ever fuse into a coherent image, it can actually help not to go back to L/R space. This technique is something of a wildcard, though; use at your own risk!)



I already made my dislike of excessive reverb clear in my earlier litany of complaints, so I'll spare you here. Still, although reverb is overused in piano recordings, it's necessary: if there is none at all, your ears will notice its absence as much as its presence. Notes cut off suddenly and jarringly; the empty spaces are disturbingly dry. Missing reverb is only vaguely bothersome in heavily pedaled music (like the first of the Three Places), though even there it can help — but in music with sharp cutoffs and crisp pedaling (such as Entropic Waltz), a lack of reverb is downright disturbing to the ears.

With reverb, I aim for that almost subliminal sweet spot where your ears notice neither reverb nor absence of reverb. I like to keep the reverb off on a side bus, with the reverb plug-in letting only the wet through, and then I can slowly bring up the gain on that bus until the reverb is at just the right level.

Reverb plugins are a matter of intense religious debate. Some are shareware. Some are horrendously expensive. Some use samples of real spaces, which often sounds very good, but gives less control over details of the sound. Religious debates surround all these options — but I find that PlatinumVerb (which comes with Logic) suits my needs just fine.

And for those of you who came looking for settings you can just copy off and use, and were all disappointed about the EQ, here's some good news for you: the settings at right are a decent starting point for any piano recording. Knock yourself out!

Gain Shaper

Gain Shaper

Piano has such a wide dynamic range, it's tempting to run it through a compressor, limiter, “sonic maximizer,” or ... well, something to make it sound louder and more saturated! In my experience, however, this is a bad idea: the shape of a piano attack has a very particular, very familiar sound, and our ears are pretty quick to detect when it's distorted.

The fundamental problem is that all these sorts of plugins work on a time scale of tens of milliseconds, which also happens to be the time scale of the initial attack of a piano hammer. I don't need adjustments that fast, however — I really just want keep the softest sections from being too soft for comfortable listening while avoiding digital clipping in the loud sections. For that, I want to adjust gain based on the sound level, like a compressor, but over a much larger time scale.

Gain Shaper automation

I wrote a plugin for just this purpose: Gain Shaper. The plugin's documentation describes what it does, so I won't explain it here. Suffice to say that it involves manually describing the shape you'd like the gain adjustments to take. I use it to make a very modest adjustment to a whole track, with the scaling set low (only 12%) so that the effect is not very noticeable — just a 2 or maaaaybe 3 dB gain in the quietest sections.

It is tempting with Gain Shaper to track every little peak and valley of sound — but it's not a good idea, because when the adjustments start happening on too small a time scale, they become unnatural-sounding, and thus more noticeable. I spent most of my time not zoomed in very far, just skimming over the tops of the attacks, in order to keep my gain changes gradual (see illustration at right). I'll zoom in close only when I need to line up a control point very precisely with a particular attack.


Peak Limiter

As I wrote above, compressors and limiters don't usually make friends with piano (if what you're aiming for is a natural piano sound, that is; they can make interesting special effects). Most of the time, therefore, I don't use a limiter at all. Instead, I carefully check for clipping by manual inspection (see “Checking the Results” below).

Occasionally, however, just a hair of limiting can be helpful — particularly when there is a hard attack next to something quiet (as in the opening of In a Perfectly Wounded Sky, for example). In these situations, I get the music as close to satisfactory as I can without the limiting, then let the limiter give me that extra 2 or 3 dB of gain.

The lookahead here is key — a limiter without lookahead probably isn't going to succeed with piano.

Level Meter Stereo Meter

Checking the Results

Although my very consistent recording setup allows me to use the same settings successfully to a very large extent with all my recordings, there is still a good deal of manual adjustment for each new piece. Most of this is related to levels — I want each piece to both:

The first of these is easy to check: watch the levels, and make sure nothing peaks above 0 dB. (I actually like to leave about 0.5dB of headroom, no matter how loud the music gets.) It's easy, however, to forget to do this: if I splice, adjust EQ, adjust reverb ... or make pretty much any change in the audio chain, I need to re-check for clipping. Furthermore, because of EQ and Gain Shaper, what looks like the peak of the piece may not actually be the loudest moment at the output. So I just leave Inspector XL's Horizontal Level Meter running, and reset the peak indicator before doing my complete final listen of the piece, just to be sure.

To check that levels are consistent with other pieces, I think about passages in existing recordings that ought to sound similar ("ah, piece X also starts mezzo piano with both pedals down"), and flip back and forth between the old finished product and the new recording. If a piece is softer overall, I might let it get a hair louder relative to a soft section in a louder piece — but for the most part, I try to keep relative recording levels similar to the relative real-life levels.

I'll also spend a minute with Inspector XL's stereo analyzer, especially if the recording seems unbalanced. Usually this doesn't lead to anything more than a slight pan adjustment. I find its "correlation" indicator nearly useless with my recordings (I'm not mixing to mono, ever, so I don't care), but its "balance" indicator is quite helpful. I don't, however, try to adjust the pan from section to section to always keep it in the center: some notes may sound very left-ish and others very right-ish, and that's just dandy by me, as long as the overall distribution is roughly even.

When everything is done, I give a final listen — no longer to details, but to the music. If the recording moves me, it's good. If something bugs me, it's back to fiddling with settings!

Next: Results >