Learn
What Is an Audio Stem?
If you work with audio long enough, someone will ask you for stems. The trouble is that the word "stems" means different things to different people, and delivering the wrong thing wastes time on both sides. This guide explains what a stem is in the strict sense, how it differs from a multitrack export or a full mixdown, and how to produce stems that are useful to whoever receives them.
What a stem is
A stem is a submix of a group of related tracks, bounced to a single audio file. If a song has twelve drum tracks (kick, snare top, snare bottom, three toms, hi-hat, ride, two overheads, a room mic, and a parallel compression bus), a drum stem is all of those tracks mixed together into one stereo file. A vocal stem might combine the lead vocal, doubles, harmonies, and ad-libs. A synth stem might contain all keyboard and synthesiser parts.
The idea is simple: take a category of sounds, mix them as a group with all their processing intact, and print the result. When you play all stems together at unity gain, they should reconstruct the full mix. That last point matters. Stems are not arbitrary groupings; they are a way of breaking a complete mix into its constituent layers, each of which carries its own internal balance, panning, and effects.
Stems, multitracks, and bounces
Three terms come up constantly in audio collaboration, and they describe three very different things.
Multitracks are the individual recordings from a session. Each microphone or virtual instrument gets its own file. A drum kit recorded with ten microphones produces ten multitrack files. Multitracks are raw material: they carry no mix decisions beyond whatever was committed during recording, and they give the recipient full control over the balance.
Stems sit in the middle. They are grouped submixes, so the recipient gets categories of sound (drums, bass, vocals, instruments) rather than every individual track. Some mix decisions are baked in, specifically the internal balance and processing within each group, but the recipient can still adjust the relative levels between groups. If the vocal stem is too loud against the drum stem, that's an easy fix. If the snare is too loud within the drum stem, it is not.
A bounce or mixdown is the final product: the full mix rendered as a single stereo (or surround) file. Everything is committed. The recipient has no ability to adjust individual elements. This is what you send to a mastering engineer in most conventional workflows, or what you upload to a streaming platform for distribution.
Understanding where each format sits on the spectrum from full flexibility to full commitment helps you choose the right one for any given situation.
Why stems exist
Stems serve several practical purposes, and the reason you are creating them shapes how you should approach the process.
Collaboration is the most common reason. When you send a mix to a mastering engineer, you typically send a single stereo bounce. But some mastering engineers prefer to receive stems so they can make targeted adjustments, perhaps taming the low end of the bass independently or adding presence to the vocals without affecting the rest of the mix. This approach, sometimes called stem mastering, gives the mastering engineer more control than a stereo bounce allows while keeping things simpler than a full multitrack session.
Remixing is another major use case. When an artist or label releases stems for remix purposes, they provide the building blocks of the track in a format that remix artists can pull apart and reassemble. Stems are preferred over multitracks here because they are easier to work with and they preserve the original mix decisions that define the song's character.
Live performance relies on stems for backing tracks. A touring act might play to a click track alongside stems for strings, pads, backing vocals, and percussion, giving the front-of-house engineer the ability to adjust the backing mix in real time without dealing with dozens of individual tracks.
Sync licensing is a particularly important use case. Music supervisors placing songs in film, television, or advertising often need stems so they can adjust levels to fit a scene. They might need to pull down the vocals during dialogue or push up the percussion for an action sequence. If you are organising project files with sync in mind, including stems alongside your final mix is worth the extra effort.
Archival rounds out the list. Full multitrack sessions can be large, complex, and dependent on specific software and plugins. A set of stems offers a practical middle ground: enough detail to reconstruct or rework the mix in the future, in a format that any DAW can open regardless of which plugins were used in the original session.
Common stem groups
There is no universal standard for how to group tracks into stems, but most popular music sessions follow a similar pattern. A typical set might include a drums and percussion stem, a bass stem, a guitars or keys stem (sometimes split further if both are prominent), a vocals stem covering leads and harmonies, and an effects or synths stem for atmospheric elements.
The exact grouping depends on the genre and the project. An orchestral score might be split into strings, brass, woodwinds, and percussion. An electronic production might separate drums, bass, leads, pads, and vocal chops. A podcast with music beds might simply need dialogue, music, and sound effects as three stems.
The guiding principle is that each stem should represent a musically coherent group, something that a recipient would reasonably want to adjust as a unit. When in doubt, ask the person who will be receiving the stems what groupings they prefer. A two-minute conversation up front can save hours of re-exporting later.
How to create stems
The process of creating stems is methodical, and getting it right requires attention to detail.
Start by identifying your groups. Go through your session and decide which tracks belong to which stem. If you are unsure, refer to the groupings above or ask the recipient. Most DAWs allow you to colour-code or tag tracks by group, which helps keep things organised.
For each group, solo all the tracks that belong to it and mute everything else. Bounce the result to a new audio file. The critical detail here is that you should leave all processing active: EQ, compression, reverb sends, delay sends, sidechain routing, and any bus processing that applies to the group. The point of a stem is to capture the sound of that group as it exists in your mix, not a dry or unprocessed version.
After you have bounced all your stems, import them into a new session and play them back together at unity gain (every fader at zero, no additional processing). The result should match your full stereo mixdown. If it does not, something has gone wrong: a track may have been left out of a group, a send may have been double-counted, or bus processing may not have been captured correctly. This verification step is not optional. It is the only way to be certain your stems are accurate.
Some DAWs have built-in stem export features that automate parts of this process, which can save time on large sessions. Even so, always verify the result by summing the stems and comparing against your reference mix.
Technical considerations
Stems should match the technical specifications of your session. If you recorded and mixed at 24-bit, 48kHz, your stems should be 24-bit, 48kHz. Do not downsample or reduce bit depth unless the recipient specifically asks for it.
File format matters too. WAV or AIFF are the standard choices for stems. Lossy formats like MP3 or AAC are not appropriate here; stems are working files, not delivery formats, and any compression artefacts introduced at the stem stage will be baked into every subsequent use of those files.
Pay attention to bus processing. If your mix has processing on the master bus (a compressor, a limiter, an EQ), you need to decide how to handle it. One approach is to remove all master bus processing before bouncing stems, so that the stems sum to a clean, unmastered mix. Another approach is to apply the master bus processing proportionally to each stem, though this is harder to do correctly and can introduce phase issues. The safest approach is to remove master bus processing, note what was on the chain, and let the recipient apply their own master processing to the summed stems.
Watch for pan and level discrepancies. A track panned hard left in your mix should remain panned hard left in its stem. If you bounce stems in mono by mistake, you lose all spatial information. Always bounce stems in stereo unless the recipient has asked for mono stems for a specific reason.
The "stems" confusion
Here is the practical reality: a large number of people in the music and audio industries use the word "stems" when they mean multitracks. A client who says "send me the stems" might want individual tracks, grouped submixes, or even just the vocal isolated from the instrumental. The word has become a catch-all, and assuming you know what someone means can lead to wasted work.
The solution is simple. When someone asks for stems, ask what they need. A quick clarification, "do you want grouped submixes or individual tracks?", takes a few seconds and can save you from bouncing everything twice. If the requester is not sure of the terminology, describe the options in plain language: "I can send you everything as separate files for each instrument, or I can group them into categories like drums, bass, vocals, and so on."
This confusion is so widespread that it is worth building the habit of clarifying every time, even with experienced collaborators. Different studios and different corners of the industry use the terms differently, and there is no authority that enforces a single definition.
Stem mastering
Stem mastering is a specific workflow in which the mastering engineer receives stems rather than a single stereo mix. This gives them more flexibility than a conventional stereo master allows. If the low end is muddy, they can address it on the bass and drum stems without affecting the vocals. If the vocals need more air, they can process them independently.
Stem mastering is not a replacement for mixing. The stems should already sound good together; the mastering engineer is making refinements, not fixing problems. If your mix needs substantial work, that work should happen at the mix stage, not be pushed downstream to mastering.
Not all mastering engineers offer stem mastering, and those who do typically charge more because the process takes longer. Whether it is worth the cost depends on the project. For a single that needs to compete at the highest level, stem mastering can provide a meaningful edge. For a demo or a work-in-progress, a conventional stereo master is usually sufficient.
Naming and delivery
Good file naming is essential when delivering stems. Each file should clearly identify the song, the stem group, and any relevant technical details. A naming convention like "SongTitle_Drums_48k24bit.wav" leaves no room for ambiguity. Avoid generic names like "Stem1.wav" or "audio_bounce_final.wav"; these are meaningless to the recipient and will cause confusion if they are working on multiple projects.
Organise your stems into a single folder, named for the song or project. Include a text file or note with the session details: tempo, key, sample rate, bit depth, and the contents of each stem. This takes two minutes and makes the package self-documenting.
For delivery, the file sizes involved in stem exports (often several gigabytes for a full song at high resolution) mean that email is not an option. Cloud storage designed for creative assets handles these transfers more reliably than general-purpose file sharing. Fabric's cloud workspace is built for this kind of media-heavy collaboration, keeping files organised and accessible without the friction of zipping, uploading, and hoping the download link does not expire.
When sharing large files, verify that the recipient can access everything before you consider the delivery complete. A quick message confirming that all files downloaded correctly and that the stems sum as expected closes the loop cleanly.
When to provide stems versus multitracks
The choice between stems and multitracks depends on what the recipient needs to do. If they need full creative control, perhaps they are remixing the track from scratch or re-mixing it in a different style, multitracks give them the most flexibility. If they need to make targeted adjustments while preserving the overall mix character, stems are the right choice. If they just need the finished product, a stereo bounce is sufficient.
In practice, many professionals keep all three available. The multitrack session lives in their DAW, stems are bounced as part of the final delivery, and the stereo mixdown is the primary deliverable. This takes more storage, but it means you are prepared for any request that comes later without having to reopen and rebounce the session.
Frequently asked questions
How many stems should a song have?
There is no fixed number. Most popular music sessions produce between four and eight stems, covering drums, bass, instruments, vocals, and possibly effects or auxiliary elements. The right number depends on the complexity of the arrangement and the needs of the recipient. Simpler productions might need only three or four stems; dense arrangements might warrant ten or more.
Do stems include effects like reverb and delay?
Yes. Stems should include all processing that applies to the group, including reverb, delay, and any other send effects. The goal is to capture the sound of each group as it exists in your mix. If you strip effects from stems, they will not sum to match your full mix, and the recipient will not hear what you intended.
Should I remove master bus processing before bouncing stems?
In most cases, yes. Master bus processing (compression, limiting, EQ) affects the entire mix and cannot be cleanly distributed across individual stems. Remove it before bouncing, and note what was on the master bus so the recipient can apply similar processing to the summed stems if they wish.
What format should stems be in?
Uncompressed formats like WAV or AIFF are standard. Use the same sample rate and bit depth as your session. Avoid lossy formats like MP3 or AAC for stems; these are working files that may be processed further, and any compression artefacts will be amplified by subsequent processing.
Can I create stems from a finished mix?
Not in the traditional sense. If all you have is a stereo mixdown, you cannot extract true stems from it. AI-based source separation tools (like those used for vocal removal) can produce rough approximations, but these are not the same as stems bounced from the original session. For proper stems, you need access to the multitrack session.
What is the difference between stems and instrumentals?
An instrumental is a version of the full mix with the vocals removed. It is a single stereo file, not a set of stems. Stems break the mix into multiple groups (drums, bass, vocals, instruments), each as a separate file. An instrumental could be created by leaving out the vocal stem, but the terms are not interchangeable.
How large are stem files typically?
File size depends on the sample rate, bit depth, and length of the song. A five-minute song at 48kHz, 24-bit stereo produces a WAV file of roughly 140 MB per stem. A set of six stems for that song would be around 840 MB. Higher sample rates and longer songs produce proportionally larger files.
Do I need special software to create stems?
No. Any DAW can produce stems. The process is simply soloing a group of tracks and bouncing the result. Some DAWs (like Logic Pro, Ableton Live, and Pro Tools) have dedicated export features that can batch-export stems more efficiently, but the basic process works the same way in any software.
What does "stems at unity" mean?
It means playing all stems together with every fader set to 0 dB (unity gain) and no additional processing. The result should match your original full mix exactly. This is the standard test for verifying that your stems are complete and accurate.
Are stems useful for podcast production?
Yes. Podcast stems typically include dialogue, music, and sound effects as separate files. This is useful when a podcast is produced in one location and mixed or mastered in another, or when content needs to be adapted for different formats (a video version might need the music bed at a different level than the audio-only version, for instance).