A stem is an audio file created by combining several tracks that share a musical role. A session might be divided into drums, bass, vocals, and other instruments. When all stems are played together from the same starting point, they should reproduce a result close to the original mix.
How stems differ from multitracks
Multitracks are the individual recording or production tracks, such as kick, snare, each guitar, and each backing vocal. Stems are submixes that combine those tracks into a smaller number of functional groups, reducing file count while preserving some control for later stages.
Where stems are used
Stems are used for collaboration, film and video mixing, live playback, remixes, and stem mastering. They also make it easier to create versions without vocals or drums and to adjust part of the balance after delivery.
Export requirements
Export every stem from the same start and end points and use the same sample rate and bit depth. Preserve leading silence when needed for alignment, and leave enough time for reverb and delay tails. Agree in advance on how master-bus processing should be handled so that recombining the files does not produce excessive level or duplicated processing.
Checks before delivery
Include the song name, part, BPM, and version in filenames where appropriate. Import all files into a new session at the same starting position, then check for missing parts, timing shifts, clipping, and differences from the intended mix.
How source separation differs
Audio estimated from a finished stereo mix by AI or other source-separation tools is also sometimes called stems. It is not the same as stems exported from original tracks: separated audio may contain leakage from other parts or audible processing artifacts.
