Mix Rescue: Stem-Mixing-Mastering

Our engineer weighs up the benefits and pitfalls of a 'stems-only' approach to mixing and mastering an album for a self-recording band.

Making smallish independent EP or album projects work commercially these days is challenging. In the '90s and early '00s, we could at least kid ourselves that we'd shift enough CDs when gigging to cover the recording and mixing costs. But while streaming services make it easier to get your music out there, they don't deliver the income that's needed in the short term. Many artists, then, view the DIY approach not only as enjoyable but also as a way to cut costs. I'm often asked to quote for mixing and mastering projects that a band has self-recorded, but while I enjoy working on such projects it can be difficult to agree a budget that works both for me and for the artist; mixing takes time, and an album project is a real commitment if you even partly rely on this work to pay the bills!

This month, I want to discuss an album project where we found a way to reduce the time I spent on the production while still adding real value. The approach sits somewhere between mixing and mastering, and it tends to work best when the artist has a clear vision of what they want from a mix — but can't quite get it there themselves. It's described variously as 'stem mastering' and 'stem mixing', and which one of those is correct really depends on your point of view, but as long as everyone's clear on the aim and approach, the name doesn't really matter!

Essentially, the idea is that the artist creates something close to a finished mix and bounces it down not to a single stereo file but as somewhere around 4-6 stereo groups, or 'stems'. They'll provide separate stems for, say, the drums, the bass, the guitars, the synths, and the vocals. (By the way, this is what a 'stem' is — stems aren't to be confused with the individual source files that you'd start with for a full mix.) Obviously, the engineer has a lot less to play with than when mixing, but still has access to much more than a mastering engineer usually would.

Inevitably, it's not a flawless approach, but it can still allow someone like me to make a significant improvement to an artist's mix in a relatively short space of time — and, importantly, for a smaller fee.

Loading In

The project was a 10-track album for rock band Derecho (see the 'Featured This Month' box). I'd helped the band with an EP a few years ago and we eventually settled on a stems approach. But this time they'd decided before recording that stem mixing/mastering would suit them. It was a straightforward project in many ways, as the tracks had been properly recorded elsewhere and the mixes were already in reasonable shape; they just needed a helping hand. But it still provides a great example of some of the pros and cons (for engineer and artist alike) of going down the stems route.

The stems for all the songs were similar in nature, with five or six 24-bit, 48kHz stereo files covering drums, bass guitar, guitars, piano/keys, and vocals. Some had a separate file for backing vocals or some sort of sound effect, and all the effects and processing were 'baked in'.

The stereo stems for a typical track from the album.You have to get into the right mindset for a project like this — to think of it more like an expanded mastering project than a full mix — and, aware that I had 10 songs to complete in a fairly short time, the first thing to do when the files arrived was work out how to set everything up in Pro Tools. When mastering, I like to A/B compare what I'm doing with where I started, but I hadn't been given stereo mixes. So the first thing I did was bounce the stems down to give me a stereo mix of each song; I'd be able to use these as references for comparison. I then loaded all 10 songs' stems into Pro Tools and went through, listening carefully for issues.

This listening stage is really important; it can be so frustrating when you get on a roll with an album project, only to realise that there's a problem with a file and you have to wait a day or two to get it sorted! Sure enough, I spotted a few little issues to flag up with the band — some noise, for instance, and a bit of clipping on a keyboard stem. This brief listen through was also important for me to develop a sense of the overall aesthetic for the project, and decide on two or three reference tracks that might help keep me on the right track. It also gave me a good idea of which songs to start work on first.

Getting Started

Most of the drum tracks benefited from some cutting in the low-mids to remove some unpleasant resonances and general 'boxiness', while Soundtoys' Sie-Q played a big part in re-shaping their tone.During that first listen, I noticed some more general issues I'd want to address. Most songs were somewhat congested in the mid-range, and I also wanted the tracks to have much more extended top and bottom ends. With the drums in particular, I could nearly always find a magical spot in the low-mids where taking quite a bit out with an EQ really opened up the sound — always somewhere between 200 and 400 Hz, and sometimes as much as a 10dB cut! I did this first with FabFilter's Pro-Q 2 EQ, but I then found myself using the Waves Scheps 73 plug-in too; I don't know quite why, but its fixed 360Hz setting often seems to work particularly well for cutting drums. A multiband compressor then held the low-frequency contribution of the kick in place, and I combined this with a narrow dip around 200Hz to rein in a slightly overbearing low note in the snare. Soundtoys' characterful Sie-Q analogue-modelling EQ plug-in was my main 'game-changer' here, though, generally applying a 2-4 dB boost at either 3.5 or 5.6 kHz, and raising the 'Low' and 'High' bands by around 3dB.

While my general approach was more akin to mastering, the amount of EQ I was using would be considered very dramatic in a mastering setting — the degree of change here was much more the sort of thing you'd expect during a mix.

As a final touch on the drums, I used Oeksound's Soothe plug-in. I'm still not entirely sure what's going on under the hood (it's more than a dynamic EQ), but it's almost magical when it comes to removing harshness! I focused it to operate on the top end of the cymbals around 14kHz, just to counter some side-effects of all the brightening EQ I'd applied.

How we arrange and process multiple guitar takes plays such a big part in achieving the desired sense of power, depth and width — and this was one area where I found it a little frustrating not to have the usual degree of control I'd enjoy when mixing.

Bass & Guitars

Getting the bass guitar sound into shape was more straightforward. Each song featured a nicely played DI track, but I felt that a little more definition would help this instrument hold its place against the guitars. To that end, I high-pass filtered the bass at around 40Hz, applied a small dip at around 280Hz, and a 3dB boost around the 2kHz area. For some songs, it also seemed appropriate to dirty things up a bit, in which case I played around with the 'Punch' and 'Crunch' settings on Pro Tools' old faithful SansAmp PSA-1 plug-in until I arrived at a setting that felt right.

FabFilter's Pro–MB, compressing just the low frequencies of the bass guitar.I knew from our previous project that this band liked quite a bass-heavy sound, so the last stage for the bass guitar was to judge carefully how far to push the bottom end. I made sure that the lower frequencies that were already present were nicely controlled, using a multiband compressor to apply somewhere around 3-4 dB of gain reduction below 200Hz. I then experimented with other areas that might benefit from a boost with something like an SSL-style EQ, and ended up with boosts ranging from 50-90 Hz, the precise range depending on how the bass part combined with the kick and the lower reaches of the guitar on each song. I also found Waves RBass effective on at least three or four tracks. This creates harmonics based on a signal's lower frequencies to give a very seductive impression of more low-end, but without making things too flabby. (I love this tool, but it's very easy to overdo it!)

Each song had 2-4 parts in the guitar stem. In a typical rock mix, of course, how we arrange and process multiple guitar takes plays such a big part in achieving the desired sense of power, depth and width — and, being frank, this was one area where I found stem mixing a little frustrating. Part of the mastering mindset for me is being able to work fast, and not overthink what I'm doing, but as different parts and solos came in and out I felt the need to adjust some of the processing and levels more than I'd have liked.

The Scheps Omni Channel was a great one-stop shop for much of the guitar processing.The Soundtoys Microshift and iZotope Ozone Imager plug-ins were used to add some extra width to the guitars.Unlike when working on the drums and bass, which were fairly quick and instinctive for me to sort out, I found myself taking quite a while just to settle on a set of tools that could do what was needed. Helpfully, I was able to do most of the processing in just one plug-in: Waves' Scheps Omni Channel. I found that a dose of saturation from its preamp section combined nicely with some VCA-style compression to add a little of the thickness I felt was needed. To get things feeling that bit more spacious, I then used the plug-in's M-S EQ to add some top end to the guitar stem's Sides signal. This, along with a little contribution from Soundtoys' Microshift and iZotope's free Imager plug-in, seemed to work nicely, the overall effect in context lending a pleasing sense of stereo width to many of the songs.

Vocals & Effects

The vocals, on the other hand, needed very little work — I liked how they'd been treated. Still, I did have to judge how bright the vocal should be for each song, and I experimented with either a conventional shelving EQ or Soundtoys Sie-Q's high-frequency control. To help control some occasional harshness around the 2-3 kHz area, I once again turned to Oeksound Soothe.

Some vocal effects had been printed on the vocal stem but I also set up a couple of effects sends of my own, and used them very subtly for the vocal and guitar stems. The first was Vallhalla DSP's VintageVerb, with a decay that varied from song to song but was always around 1s. The other was a slap-style delay, which brought a little extra depth and dimension to both the vocals and the guitars.

Bus Processing & Mastering

It's important to point out that the processing I've described so far was done with some stereo bus-processing plug-ins in place from quite an early stage, a tactic that's often referred to as 'top-down' mixing, and you'd certainly notice the difference if you removed the bus processing, especially the EQ!

The Softube Tape plug-in helped glue the low-end of the mix together.Most of the mix-bus processing was done via the EQ section of the Big Ceil plug-in by Acustica Audio.First in this chain was Softube's Tape plug-in, an analogue-modelling open-reel tape–machine emulation, which seems to do that subtle and difficult–to–describe thing that analogue tape does, whilst helping glue the low end of a mix together. This was followed by some SSL-ish bus compression, but the real action came courtesy of Acustica Audio's Big Ceil, a plug–in that uses the company's proprietary convolution–based technology. This slightly mysterious channel strip plug-in isn't generally what I'd call subtle-sounding, but its high- and low-shelving EQ filters seem to do an excellent job of opening up the top and bottom ends of a mix or master.

Once I was happy with how each song was sounding, I wanted to get the songs in a more traditional mastering setting and think about the project more as a whole. I love PreSonus' Studio One for this part of the process, so after I'd bounced all my mixes from Pro Tools, I laid out the album in Studio One and focused on getting the 10 songs to sit together nicely as an album.

The mastering stage and album sequencing were done in PreSonus Studio One Professional.

The sequence of the tracks had already been decided by the band, so a large part of this was about judging the relative levels of the different songs. To make that process easier, I'd been very careful when I'd bounced the mixes down to leave myself a bit of headroom to play with. Once I'd decided how loud the loudest parts of the album were going to be, I deliberately treated this stage in a very un-technical way — I just listened and made judgments about how the songs flowed into one another, tweaking accordingly. I also checked that the loudest sections at different points in the album felt like they were at or around the same level, and listened closely to make sure there was a broad tonal consistency across the album; I found myself making very small changes with low- or high-shelving EQ, and generally doing a few nips and tucks.

Blurred Lines

I was really pleased with the results overall, but as the difference between my results and the mixes I'd bounced at the outset was not subtle, I checked how they compared with some reference tracks before, broadly satisfied, I sent everything to the band. The response I got wasn't perfect...

Don't get me wrong, it was certainly very positive and constructive, but it also highlighted some of the pitfalls of choosing stem mastering in place of mixing. I'd been careful not to change the instrument levels in each song too much, but I had changed the overall tonal balance considerably. You have to use your judgement here but, inevitably, such changes can lead an artist to see their song in a different light! The band wanted me generally to 'tilt' the tone of the album, to give it a slightly less bright and more bass-heavy sound, and that was easy to address... but they also requested some detailed changes to the levels in certain parts of songs.

While happy enough to do the extra work, it started to morph more and more into a traditional mixing job — and it wasn't easy. Much of the work involved the guitars and, on reflection, it might have been better to have some of the busier tracks split out into a few different stems, to give me a little more control; I'll certainly be thinking about that next time I'm auditioning the stems for a project like this. But this is where the whole concept of 'stem mastering' can be difficult to pin down: just how far do you want to travel down this road of requesting more stems before you lose all the time and cost benefits that led you to agree on the stems approach in the first place?

Summing Up

If you're an artist who wants to try stem mixing/mastering, it's important that your aims are clear and realistic, and I'd suggest that you test the water with a song or two to make sure the approach works for both you and the engineer. But it certainly can be worthwhile. A good mastering engineer can work wonders with a stereo mix, but there's a limit to what they can do with a stereo file, and I think that for self-producing artists — especially those who want a degree of control over the mixing but don't have the experience or monitoring environment to achieve professional-sounding mixes — this middle ground could potentially be a great way to finish a project.

As an engineer, I quite enjoy the limitations stem mastering imposes on me. For one thing, it's easier to resist going down the rabbit hole in search of, say, the perfect kick sound, if the hole isn't there in the first place! And there's a lot to be gained from working quickly, with broader strokes too. The bottom line is that it enabled me to make a significant contribution to this project in maybe a quarter of the time it would have taken me to mix from the raw multitracks.