Cleaner, Easier Stem Isolation with Music Rebalance in RX

Music Rebalance in  RX  uses processing trained with neural nets to individually identify, separate, and adjust the gain of four critical groups of mix elements: vocals, bass, percussion, and other instruments. It gives you more control over a mix so you can make your own creative choices for mixing and remixing, without needing stems or instrumentals.

Since you don’t need stems, Music Rebalance is perfect for adjusting older recordings, or isolating single elements (like vocals!) No need to waste time going back into your session, loading up a bunch of stems, and then soloing out the instrument you want to remove. Watch the video below to learn more about Music Rebalance.

This article references a previous version of RX. Learn about RX 10 and its powerful new features like Dynamic Adaptive Mode in RX De-hum, improved Spectral Recovery, the new Repair Assistant, and more. 

Learn More

Let’s explore some of the improvements to Music Rebalance in RX 8. I recently used it on a jazz project with loud drums and a comparatively soft organ bass. Here’s the original, followed by the Music Rebalanced version.

Now it’s time to explore how some of the new features in Music Rebalance helped to achieve this result.

1. ARA 2 support for Logic Pro X 

Now within the Music Rebalance module, Audio Random Access 2, or ARA 2, allows Logic Pro X to exchange more data with RX than standard formats. Unleash the full power of Music Rebalance without having to leave your session, and maximize efficiency in your production. Check out the video below to learn more.

2. Solo buttons

Without hearing a single second of audio, here’s an improvement right off the bat: you can now solo at will!  Say you wanted to rebalance a stereo file, and you wanted to audition each isolated element. In RX 7, you’d have to keep notes. With RX 8, that’s no longer necessary thanks to the new Solo buttons in Music Rebalance!

3. Better isolation algorithms

The original Music Rebalance was already top-notch—it hung with, if not trounced, all similarly-priced competition. The tool in RX 8 is even better. To show you exactly how much better RX 8 can be, let me present the file I’ll be working with:

Now, here are some drums I grabbed with RX 7’s Music Rebalance on its highest sensitivity first, and then some drums I grabbed with the new RX 8 Music Rebalance at ¾ sensitivity. 

I’ve only gone 75% to the max with the Separation slider offered in RX 8, and yet, the separation is much more powerful: we hear far less of the surrounding instruments—the guitars and bass.

4. Better separation slider and GUI

We can see some cosmetic differences between Music Rebalance in RX 7 and RX 8 that are quite helpful: RX 7 offers a Sensitivity slider, adjustable for each individual instrument. RX 7 also provides a dropdown menu for various algorithms—Channel Independent, Joint Channel, and Advanced Joint Channel.

We’ve lost independent sensitivity controls in RX 8. Instead, they’ve been replaced with a global Separation slider, which ensures that each stem will be separated with the same parameters. Also, instead of picking between Separation algorithms, we now see three more intuitive options in the Quality drop-down menu: Good, Better, and Best.

5. In-module stem separation

One of the big headlines from RX 8 is the ability to separate stems in one simple step. Simply click on the Separate button, and RX 8 does the rest.

We can export our resultant files and bring them into our DAW—undeniably useful in a variety of circumstances.

6. Better stem phasing

Now maybe I’m a huge dork, but this gets me really excited: RX 8 creates stems that null to a much greater degree with the original track.

What does this mean? When I line up the original file with the resulting, separating stems, I can flip the polarity of the original track, and I’m left with a file that is barely audible, and measures a measly -65 LUFS in the meters.

Okay, but why is this important? 

Simply put, null tests constitute incontrovertible evidence that a track has been manipulated or affected. If I take a track, copy it, and flip the polarity, it will give me silence when I hit play—the files will null. If I don’t have silence, I know the two tracks are different. This seems pretty obvious, until we get into degrees of nulling.

If I line up an affected track with its original source, and I get only a whisper of noise, that tells me the differences are minimal. In this case, playing generated RX 8 stems against the original file gets me a loudness measurement of -65 LUFS, integrated. And to be sure, this is not silence.

But -65 LUFS is pretty damn quiet. And it’s a hell of a lot quieter than, say, -30 LUFS. I wonder why I pulled that number out of thin air…

Spoiler: I didn’t. A null test with stems rendered from RX 7 gives us a file of -30 LUFS integrated:

But in RX 8, the process clearly introduces fewer—and quieter—artifacts, as shown by the test:

7. All these things are a boon to mix and mastering engineers

We have built a foundation with these five facets of Music Rebalance, and on this foundation we can build a house.

Because we can audition easily in solo, because the GUI is simpler, because the isolation algorithms are better, because we can export stems with a click of a button, and because these stems null to a good degree, we have unprecedented power in dealing with unruly material. Indeed, we can attempt previously-impossible audio feats.

Take this mix again—excerpted here:

The song is an unreleased, upcoming tune by Leland Sundries called “Party Bus,” and there are things I really like about this mix: it has heart, grit, drive, power, and emotion. It also has a classic garage sound that’s ever so much fun.

But there are some things that I would like to improve—things that are hard to improve without either a remix, or settling for compromise. The high-hats are quite bright, but if I tamp them down, the voice will suffer. The voice is a bit spitty, but if I de-ess, the guitars will suffer. I’d like to bring up the volume of the bass, but this will affect the kick drum; it will also get in the way of the guitars.

Most glaringly to my ears, the lead vocal is not centered. Maybe that’s a choice—it’s a valid one if done with intention, but if not, I don’t have the power to change it. If I try to broadly rebalance the stereo file to position the vocal in the center, the kick, snare, and bass will not.

Without a remix, compromise would be the only way:

It’s a bit too beefy, it’s a bit too shelved in the highs, and the vocals are still off-center.

Enter Music Rebalance in  RX : separate this mix into stems, and we can do a lot more. We can reposition the lead vocal, de-ess it a little, turn up the bass, and de-harsh the high-hats, all without affecting too many other elements.

Takeaways

We must deliver our standard warning: Music Rebalance in RX is a powerful tool, and in the wrong hands, it could generate devilishly weird results. It’s arguable, from a purist’s standpoint, that I might’ve gone too far in demonstrating its capabilities with that last example: certainly, it’s better to talk to the production team about their intentions than to pull their mix apart at the seams.

Still, if you’re careful, and if you’re judicious about it, you can achieve unprecedented feats of audio with this new version of Music Rebalance. It’s a balancing act—one that we leave entirely in your hands.