Stem (Audio Stem)

A finished song may have dozens or even hundreds of tracks. Sharing all of them is unwieldy, and sharing only the final stereo mix allows no changes. Stems sit in between: the tracks are grouped by role, and each group is exported ("bounced") as a single audio file. A typical set might be drums, percussion, bass, guitars, keys, lead vocal, backing vocals and effects.

What stems are used for

• Mastering: some engineers prefer stems, so they can adjust, say, the vocal against the band (mastering).
• Remixes: a remixer can replace the drums but keep the vocal.
• Film, TV and games: a mix engineer can lower the music under dialogue, or drop the vocal out.
• Live shows: backing tracks are often played as stems, so the sound engineer can balance them in the room.
• Practice: muting one stem gives you a play-along track. Mute the bass stem, for example, and play the bass part yourself.

Exporting stems properly

1. Start every stem at the same point, usually bar 1 or the very start of the project, even if the instrument enters much later. Then every file lines up when the stems are dropped into a new session at zero.
2. Make every stem the same length, through to the very end of the song, including reverb and delay tails.
3. Use the session's sample rate and bit depth, and name each file clearly, for example "Songname_Drums".
4. Decide about effects. Stems usually include their own processing; shared reverbs and delays can be printed into each stem or delivered as a separate effects stem. Say which in a note.
5. Check the result. With every stem playing at its exported level, they should sound like the original mix, apart from any processing that was on the final mix output.

Stems, multitracks and stem separation

Multitracks are every individual recorded track, unmixed; stems are groups of them, already mixed within each group (multitrack recording).

Stem separation is something else: software that analyzes a finished stereo recording and estimates the vocals, drums, bass and other parts from it. It can be very useful for practice and transcription, but the results can contain audible artifacts, and separating someone else's recording doesn't give you the right to release the parts (sampling).