What Are Stems in Music? The Complete Guide

Stems vs multitracks vs tracks vs the master

This is where most of the confusion lives, and it matters because asking a client for "the stems" when you need the multitracks will cost you a day. Here's the whole hierarchy in one view.

So a track is the smallest unit and the master is the largest. Stems sit one level below the master: fewer files than multitracks, each already balanced and processed the way the mix engineer intended. Multitracks are raw and granular — you'd hand those over only when someone needs to rebuild the mix completely.

Why stems exist — what engineers actually use them for

Stems aren't busywork. They're the format you reach for whenever the balance between parts needs to stay editable after the mix is done:

• Stem mastering — hand a mastering engineer 4–6 stems instead of one stereo file and they can nudge the vocal up or tame the low end without you re-bouncing the whole mix.
• Remixing and remix packs — a remixer needs the vocal, drums, and bass as separate files to build something new; nobody remixes from a flattened master.
• Sync licensing — film and TV libraries require stems so an editor can duck the vocal under dialogue or drop the drums for a tense scene, all from your one cue.
• DJ sets and live remixing — stem-capable software (Serato, Traktor, rekordbox) splits tracks into drums, bass, melody, and vocals so DJs can mash parts in real time.
• Live backing tracks — a touring act runs stems to a multi-output interface so front-of-house can mix the band against the playback every night.
• Archiving and future-proofing — stems are a compact, session-agnostic snapshot; you can reopen a song in any DAW in ten years without chasing missing plugins.
• Post and podcast delivery — splitting dialogue, music, and effects into stems lets a broadcaster meet loudness specs and swap a music bed without a re-edit.

How many stems are in a song?

There's no fixed number — it depends on who's receiving them and why — but most songs reduce cleanly to four or five groups. This is the standard breakdown, and what each group usually contains:

Scale up or down from there. A sparse singer-songwriter track might be three stems; a dense pop production handed to a sync library might be ten or more, with vocals split into lead, backing, and ad-libs and a separate instrumental and a cappella. When in doubt, ask the recipient for their deliverables list — sync libraries and mastering engineers usually have one.

What format are stems delivered in?

Stems are delivered as lossless WAV or AIFF, never MP3. Match the session: if you tracked at 48 kHz / 24-bit, your stems are 48 kHz / 24-bit. The whole point of a stem is that the next person processes it further — mastering, re-balancing, time-stretching — and lossy compression throws away exactly the detail that processing exposes. An MP3 stem is a corrupted starting point, and you can't un-bake it.

How to export stems correctly

A clean stem set lines up instantly and sounds identical to the mix. Most stem problems are avoidable; this is the checklist that prevents the email that starts with "these don't line up":

1. Bounce every stem from the same start and to the same length. Start at bar 1 (or the session start) and end at the same point for all of them, so they stack in sync no matter where they're dropped.
2. Leave the effect tails. Let reverb and delay ring out past the last note; cutting the export short clips the tail and the stem won't match the mix.
3. Match the session's sample rate and bit depth — no resampling, no dithering down to 16-bit unless you're explicitly asked.
4. Keep faders, mutes, and automation exactly as in the final mix. A stem is a snapshot of the mix, not a fresh balance.
5. Decide on master-bus processing deliberately. For stem mastering, bounce with your master limiter and bus compressor bypassed and leave headroom; if the client wants a printed sound, deliver both a processed and a clean set.
6. Name them clearly and in order — 01_Drums.wav, 02_Bass.wav, 03_Vocals.wav — then zip the folder so nothing gets lost or reordered.

Aim for around -6 dBFS of peak headroom on stems destined for mastering. That gives the mastering engineer room to work and keeps any single stem from clipping when they're summed. Check it on the way out instead of guessing:

Stem mastering — a master from just the stems

Stem mastering is mastering from a small number of stems instead of a single stereo file. It sits between mixing and mastering: the engineer can correct balance issues a stereo master can't touch — pull a hot vocal back, add weight to the bass, widen the music — while still doing the loudness, tonal, and dynamics work of a normal master. It costs more and takes longer than stereo mastering, but it's the move when the mix is close but not quite locked. If you're prepping a stem master yourself, set your targets first:

Delivering stems to a client without the mess

The exports are the easy part. The mess is everything after: a zip in a chat thread, a re-send because the vocal was named wrong, a client who can't tell the v3 stems from v2, and an mp3 "preview" floating around before anyone's paid. Stems are big, they get versioned, and they often gate a payment — which is exactly the workflow a delivery Room is built for: labelled files, a clean preview the client can hear, and the full-quality WAVs unlocked the moment the invoice clears.