How to Export Stems Correctly

Most revision rounds on a mix are not caused by taste disagreements. They are caused by a stem export that has already thrown away information, or that makes the engineer rebuild something that shouldn't have needed rebuilding.

This guide covers the export settings and the file hygiene that avoid it.

The five settings that matter

1. Consistent start point. Every stem must begin at bar 1 — the same zero point — including any silence before the first note. If stems start at different times, the engineer cannot align them without guessing.

Leave one to two seconds of silence before the first downbeat across all stems. This makes the alignment unambiguous and gives reverb and other processing room to resolve at the head.

2. Bypass all master bus processing. Limiting, glue compression, stereo widening and saturation on the master bus must be off.

The reason is not that these are bad — it's that they cannot be undone. A limiter that has already shaved your transients has removed information. A master-bus widener has already altered your stereo image. If you send a mixdown-shaped export, the engineer is working on a degraded copy of your song, not the song.

Keep your master-bus chain in the session so you can show it to us. Include one stereo "rough mix" export with that chain, clearly labelled — it's a useful reference. Just don't send it as the source.

3. No clipping, and sensible headroom. Peaks should stay below 0 dBFS with a few dB of margin. If the export shows inter-sample peaks or flat-topped waveforms, the damage is permanent and will be baked into the final record.

Aim for peaks around −6 dBFS. There is no benefit to being hotter and a real cost to being at the ceiling.

4. Consistent sample rate and bit depth. Export every stem at the same sample rate as the session — don't resample individual files. 24-bit is the standard working depth. Do not dither individual stems.

5. Mono for mono sources. If a source is mono, export it as mono, not as a duplicated stereo file. A mono vocal exported as identical L/R loses nothing but wastes space, and it makes panning decisions clumsier later.

Once the files are exported, run them through the stem checker before you send the set. It reads each file in the browser and lists sample rate, bit depth, length, peak and leading silence side by side, so the settings above are confirmed rather than assumed. Nothing is uploaded — the files stay in the tab.

What counts as a stem

There is a real difference between stems and track-outs, and it matters for what a mixer can do.

Stems are small groups — drums, bass, music, vocals. Good for archival and for live playback. Too coarse for mixing, because a balance decision inside a stem is impossible if both elements are baked together.
Track-outs are one file per original track, with anything sharing a channel printed together. This is what a mixing engineer actually needs.

Send track-outs. If you also send stems, label them clearly so they aren't mistaken for the source.

Not sure the export settings are right yet? Run the stem checker first, then send the files.

Naming and the reference package

Name files so the engineer can understand the song without opening the session:

01_KICK.wav

02_SNARE.wav

03_HAT_MAIN.wav

04_BASS_DI.wav

05_LEAD_VOX.wav

06_LEAD_VOX_DOUBLE.wav

07_ADLIB_MAIN.wav

08_SYNTH_PAD_wide.wav

Alongside the audio, include:

A rough mix. Your own balance, however rough. It tells the engineer what you were aiming at, and it's the fastest way to communicate something that is difficult to describe in words.

Two or three reference tracks. Songs that sound the way you want yours to sound. Note specifically what you like about each — the low end, the vocal tone, the width. A reference without a note is half useful.

The technical facts. Tempo, key, sample rate, and anything unusual (tempo changes, a section recorded at a different session, a part that is intentionally out of time).

A note on what you want to keep. If a section is meant to be raw, say so. Engineers default to polishing, and without guidance you may get a cleaner version of something that was supposed to stay rough.

The one thing that causes the most rework

A single combined export of everything — vocals plus instrumental, or worse, a stereo mixdown labelled as stems. If the engineer cannot change the vocal level independently, the mix cannot be genuinely mixed. It can only be processed.

If you're unsure whether your export is correct, send the file list and we'll check it before you upload. That's cheaper than a revision round.

This guide is general audio engineering knowledge, written for artists. If you'd rather not deal with any of this, we can accept a full session archive instead of individual exports.