We’ve all heard of stems but what are they?

I would like to clear up a few misconceptions about stems because a lot of people, including A&Rs, music supervisors and post-production supervisors use the term incorrectly. Stems are now commonplace in music, and indeed many other areas of audio production, but not just as a delivery requirement: creating stems in your own projects can make for convenient and time-saving workflows, both when working with others and when making music on your own.

As such, we really must all be on the same page when we’re discussing them. 

What Stems Are

Stems are sub mixes of a larger mix, that when played together at equal volume will exactly recreate the full mix. 

Typically, this will mean dividing larger ensembles, whether recorded naturally or artificially, into smaller subsections. An orchestral mix might be stemmed into strings, brass, woodwind and percussion, or a rock band into drums, guitars, keys and vocals. Crucially, stems are groups of elements that make up the final mix, not the individual elements themselves.

As such, anybody given the stems can alter the relative balance of these sections. For instance they’ll be able to control the levels of the brass against the strings, or the guitars against the drums, but not the individual instruments such as violins against the violas, or the kick drum against the snare drum. 

Stems are used in post-production too. The most common stems to come out of a post mix. It starts with Dialog, Music and Effects, often shortened to DME. The DME is often made up of Dialog (or sync sound for documentaries) Narration or Voiceover, Music, Effects, although larger projects can certainly go wider, with additional stems for foley, backgrounds and diagetic music (music in the scene playing through a radio or a PA system), or any other group of similar sounds. But the focus of this article is the use of stems in music production.

Stems are most useful for allowing somebody further up the chain of delivery to have some control over the mix, but not complete control. 

Please note, I’m talking about audio files here. Although whilst mixing it’s common to group tracks together and route to a bus to simplify things or add bus processing, stems are what we’d get if we bounced these groups as separate files. The stem is an audio file, not the auxiliary channel itself within a DAW project. 

Multitracks Or Stems? - What Is The Difference

Before I go any further, I want to clear up what stems are not. It is a commonly held misconception that stems and multitracks are the same thing and in recent years I’ve really noticed that many are incorrectly using the terms multitracks and stems interchangeably.

In the very first YouTube video I watched whilst researching this article the vlogger tried to explain what stems are and they just described multitracks - in fact, one of the comments was “so stems are just multitracks” and the author actually liked the comment!

So what is the difference? If multitracks is also an odd word for you, just think instead of “audio files”. The term multitracks comes from multitrack tape - a single piece of tape that could hold multiple, well, tracks. Much like stereo tape can hold two tracks of left and right, multitrack tape can simultaneously record and playback up to 24 tracks across 1 reel of tape up to 2 inches wide in the most common instance. Although physical tape as a recording medium is now generally long gone, the term multitracks is still used to describe all of the audio files that make up your recording and will be mixed into the final piece of music. 

I’m not explaining this because I’m a nostalgic tape nerd: If you’ve only ever worked in a DAW perhaps you’ve never actually come across the term multitracks, I can understand why you might confuse multitracks with stems. 

Here is a useful chart to summarise the differences between stems and multitracks:

An important single element and its effects might have a dedicated stem. For example, in an orchestral mix the piano might have its own stem, or in a rock mix the lead vocal might have its own stem; however, these stems will still include reverbs and other effects, so they are still a combination of elements. 

Problems When Confusing Stems And Multitracks

The confusion between stems and multitracks is oftentimes harmless, but I’m not just being pedantic like when I point out to someone that they’ve said “literally” when they mean the opposite of literally.

Unfortunately, the mixing up of the terms stems and multitracks has led to problems a few times in my career. 

A few years ago I recorded a single and a B-side for a band. The single went off to the mixing engineer, but my monitor mix of the B-side was going straight to the mastering engineer, neither of whom I knew. The band said that the mix engineer has asked for the stems - I said this was fine; they also said that the mastering engineer had asked for stems - I said there weren’t any. And for a moment I looked like a massive, uncooperative dickhead. 

Of course, I realised that in the first instance the mix engineer (or perhaps the band, passing on the question) didn’t mean stems but rather multitracks, whereas the mastering engineer was probably just inquiring as to whether the mix had been stemmed out, which it hadn’t. Everything was fine, but if we hadn’t had the conversation in person I don’t think it would have been a good look. 

In a more worrying example, I’ve also been in a couple of overdub sessions where we’ve been told that we’d be tracking to stems, only to receive a massive multitrack Pro Tools session with dozens of missing plugins two minutes before the recording was due to begin.

Finally, although this hasn’t happened to me, imagine being a live sound engineer waiting for stems of the backing tracks only to receive multitracks a few minutes before the doors open. I’m sure they’d be pretty annoyed. 

Due to these and probably many other real-world problems, please: make sure you know the difference between stems and multitracks. 

Why Stems Are Useful When Collaborating?

Although this might have cleared things up for you, you may now be wondering “what’s the point?”. After all, when you’re happy with a mix that has taken you a long time to finish, why would you want to create extra work for yourself by dividing it into a bunch of submixes for somebody else to mess around with and potentially ruin?

Music Scoring For Film And TV Productions

Stems are most common in score mixing for film and TV. Delivering music stems to the dub (the final post-production sound mix) gives the dub engineer / re-recording mixer the option to re-balance the music within the context of overall soundscape. 

In the simplest and most common instance, if the music is interfering with the dialogue then the dub engineer can simply turn down the stem that contains the conflicting sound, rather than being forced to turn down the music as a whole.

Let’s imagine that the dub engineer is mixing an action scene with a lot of big explosions in the sound effects and a wonderfully written score that features huge, exciting drums and a large brass section playing the hero’s theme. Unfortunately, the explosions and the drums are occupying similar frequency bands, fighting for the low end and everything is sounding horrible. 

Without stems, the dub engineer would have to choose between either the music or the effects: they could either turn the sound effects down to let those big drums through, or they could turn the music down to make room for those big explosions - but in the latter case they’d also be turning down that wonderful brass line.

However, if the brass and the drums are on separate stems then the dub engineer can tweak the music mix slightly by bringing the percussion stem down a little, maybe only for a few moments at a time, so that they don’t interfere as much with the explosions. This way, the brass melody is still shining through at its intended volume.

Or perhaps in an animated adventure when the snooty cat and the courageous dog with the celebrity voices meet for the first time, there’s some lovely orchestral underscore. Unfortunately, at the crucial moment, a flute trill is really getting in the way of the dialogue. It would be a shame to have to turn the whole music track down because otherwise the strings, brass and percussion are working really well. 

Luckily, there is a separate woodwind stem, so the dub engineer need only pull that particular stem down at that point, whilst retaining the levels of the rest of the orchestra, and the scene works out perfectly.

You may be thinking here that surely it’s a composer’s job to write music that fits the picture? Indeed, it is, but anyone who’s worked on a film score in recent years will know that the picture can change from day to day, and certainly long after the music has been recorded and mixed. In which case, the score might not fit the new picture as well as the picture it was written to. Stems are useful for the dub engineer, as well as to a music editor who may have to make drastic edits to make the wonderfully written music fit the new edit. 

Although composers are sometimes reluctant to allow a dubbing engineer or music editor to alter their work, stems are almost always required as deliverables. If there is to be a separate music release of the score, the composer will usually be given creative control and the full mixes or stereo fold downs from the score mix will be used.

You may also be wondering why we don’t just give the multitracks to the dub engineer in the first place? Well, we want the dub engineer to have some control over the mix but not so much that they can change the very nature of the score, it’s more of a re-balance rather than a re-mix. Secondly, in reality, the dub engineer doesn’t want that much control either - they already have enough on their hands.

Stereo Music Mixing

The idea that stems can be used to make collaboration in scores easier can be extended to more general stereo music mixing as well.

Sometimes, mastering engineers like to have stems. As the final pair of ears that most projects usually need, they might think for instance that the backing vocals need bringing up just a tiny bit, or the keys are maybe just a little too loud. With the stems, they don’t need to suggest a remix, but instead, make those little re-balancing tweaks themselves. Or perhaps the EQ’ing sounds great on almost everything, but it makes the guitars just a little bright. With stems, they can leave the guitars out of their signal flow. 

Blending the boundaries between mixing and mastering, perhaps a producer is ready for a song to be mixed but is already happy with the balances between certain groups of instruments. As such, instead of giving a mix engineer all the multi tracks, they might prefer to give them a number of stems to work with. I’ve found that this approach of “mix from stems” has become much more common in recent years, as home producers and engineers become more confident and able but still want to benefit from a professional mix engineer and a commercial studio’s outboard gear. 

Stems can be useful midway through a project too. If you are recording an overdub at another studio, rather than send a whole multitrack session to a recording engineer using a different setup, and who is unfamiliar with the project, it would be easier if they can instead work with a small number of stems. That way if the performer would like a bit more or less of something in the headphones, it’s much faster to adjust a well labelled stem than to search for a bunch of tracks in a complicated DAW session. 

This could also apply when sending a player something audio to practise to, or an orchestrator who needs to work out parts and would like to hear certain sections of the mix in a bit more detail.

Stems For The Lone Producer

So far, we have made a generalisation that stems are most useful for passing a project on, but even if you don’t collaborate with others, how might stems be useful for you?

Those who assisted in studios on any pop sessions may remember the dreaded vocal up, vocal down, and instrumental versions that were generally required - the engineer would go home at a sensible time and the assistant would have to stay behind and print numerous versions before they could unplug the patch bay and zero the desk ready for the next session (usually remembering half way through that they should have done a recall and going into a blind panic).

In the age of faster than real-time bouncing, creating versions might not be the huge time sap that it once was, but on big projects with hundreds of processor hungry plugins a bounce can still take a while. Even if you only create the limited stems of instrumental and vocal, by bringing these two files into a new DAW project and bouncing versions from these, the bounce will be much, much faster.

Furthermore, if two weeks after you’ve delivered the final mix the record company suddenly wants versions without the BVs and without the solos, bouncing these from stems can be done quickly and simply, even on a computer with limited resources and without the same plugins.

Elaborating on this further, if you need to revisit a particularly large mix on a less capable system, if you’ve printed stems you may only need to work from these to make a tweak - a relatively broad request such as “raise the guitars” will be really straightforward on even the most basic system. 

We all know those clients who assume that because they’re working then you’re working. Perhaps they call you in London at 10pm in the middle of the mastering session in LA at 2pm, saying the mix is perfect except that the backing vocals need to go up a tiny bit in the final chorus, and they need the change right now. Rather than returning to your main workspace in the middle of the night, if you have access to the stems you can run this off on a laptop in a couple of minutes and look like a hero!

Or let’s say that the client wants a change that is a little more complex - perhaps adjusting the balance between two instruments that share a stem. If working in the box then this isn’t really much of a problem, but if you’re using outboard or even a whole analogue desk, at least if you printed stems then you’re only going to need to recall and tweak that single stem as opposed to the whole mix. 

Stems can also be used as a kind of archiving. Maybe one day you’ll want to open up a session from a few years back, only to find that an essential plugin is now longer supported. Fingers crossed this won’t matter if you bounced it all to stems. 

Finally, of course, one day it might be a delivery requirement to provide stems. 

How To Make Stems

Stem Creation - Multiple Passes In Solo

There are a few ways to create stems. The first method is to simply solo the relevant groups of tracks and bounce them out one stem at a time. This method has the advantage of allowing you to not worry about the stems until it comes time to making them, and if you have a speedy computer that can bounce much faster than realtime then perhaps this method will work for you.

Unfortunately this method is time consuming, and it can be very easy to accidentally miss tracks out. As such, it’s a good idea to compare the summed stems with the full mix, and perhaps even flip the polarity of the full mix against the stems to double check that they’re cancelling out. If not, then you’ve missed something (although it should be noted that some algorithmic reverbs won’t cancel out as they produce very slightly different results every time).

Stem Creation - Single Pass, Multiple Stems

However, I prefer to print my stems in realtime in a single pass. This is partially out of habit - remember that even in-the-box mixes in Pro Tools had to be bounced in real time until version 11, so I’m just used to listening to a mix in full one final time as it goes down which is a helpful quality control. Also, I cut my teeth assisting on film score sessions mixed on a big analogue desk which was only capable of printing four 5.1 stems simultaneously. If more stems were needed, we had to do a dreaded second pass which, after sorting out the naming and routing would actually mean it would take longer than twice the amount of time. This was not what we wanted to do at 3am.  

To print stems quickly, instead of routing everything to the main output, instead route each group of instruments that you want to make into a stem to a new audio track. The outputs of these should either be your main output, or a full mix audio track that is routed to your main output. This way, you can record your stems and full mix directly to audio tracks within your DAW session.

More Stems, More Effects

The main thing to bear in mind is that you’ll need to duplicate some resources here. On a regular mix you only need one of every effect, say reverb and one delay. But when stemming you need one of these for every stem, routed to the relevant stem bus. Otherwise, you’d have the effects of all the different stems on one stem, and the point is to separate things. So if you’re creating four stems you’ll need four sets of effects busses. You can imagine how quickly this will start to take up system resources if you’re printing a lot of stems, and especially if you’re working in 5.1 or 7.1

Finally, the eagle eyed among you will have realised that you can’t have a master bus compressor when delivering stems. Because dynamic effects react to the level you put into them, compressing each stem individually will not have the same result as compressing the sum of all the stems. 

So, if you mix with a bus compressor, unfortunately you can’t bounce your stems through it and expect the final result of the summed stems to be equal to your final mix. On the other hand, frequency based effects work irrespective of the input level, so you can copy some sweetening EQ to every stem.

I’m afraid there is not much you can do about this. You can, of course, apply bus compression after the stems in the signal chain, but then you need to be able to guarantee that whoever receives the stems can replicate this setup, which for my liking is leaving too much to chance. A better approach is use a buss compressor over each individual stem that you tweak as you mix. 

(That all said, with a complicated setup of side-chains and busses, you can sort of replicate buss compression on individual stems, but it’s cumbersome, has limitations and is beyond the scope of this article!)

Some Thoughts On Stem Delivery In Music Production

In my opinion, when delivering stems consistency is key, because differences between stems could indicate a mistake. There are a couple of good practises for when you’re making your stems to avoid this. 

Firstly, make sure that your stems are all the same length. If you’re given a bunch of stems, you’d probably assume that they all start at the same place, so dragging them all to the beginning of your DAW session would mean they’d be playing in sync. All well and good.

But whenever I receive stems of different lengths doubt immediately creeps into my mind. How do I know that one of the shorter stems doesn’t actually have some crucial audio towards the end that’s been missed off by mistake? Or similarly, maybe something at the beginning was missed, and the shorter stem was in fact accidentally bounced from midway through the song, meaning I’m placing the stem at the wrong start point and it’s playing too early? Or that one of the stems has accidentally included a few extra bars of count off and is playing too late? Or perhaps a colossal mistake has been made, and all the stems actually start in completely different places? Or maybe it wasn’t a colossal mistake, but each stem needs to be spotted into place at a particular bar, and this was all written out in a spreadsheet for you - but you never got the email?

These are all very possible and easy mistakes to make whilst making stems - especially when they’re being bounced out late at night ready for a session the next morning. A way to remove any doubt from a recipient’s mind is to bounce all of your stems at exactly the same length. You’ll immediately be able to see that both the beginnings and the ends of the files are lining up exactly with each other, instantly reassuring you that everything is in sync. 

Another inconsistency to avoid is delivering stems with different channel counts. This was more problematic in the days when Pro Tools could not work with interleaved files, because receiving four 5.1 stems and a single LCR stem would make me think “hang on, where’s the rest?”

But even today, receiving some files in 5.1 and others in stereo makes me wonder if an upmixer was set incorrectly on one of the stems, or if someone has accidentally sent me a reference stereo folddown by mistake. Again, you can remove any doubt from a recipient's mind by keeping every format consistent throughout. 

If you think that doing either of these things is actually beneficial because you're saving disk space, remember that disk space is not the expensive commodity that it once was. A few extra megabytes of digital silence in your audio is not going to save you any money, but it might save you from receiving a phone call from a confused audio engineer.