Preparing Stems for Mastering - A Client Guide

There is no single universal standard, but two approaches come up again and again.

The full breakdown (up to ~8 stems). Most mastering engineers are happy to work with anywhere up to eight stereo stems. A typical breakdown looks like:

• Drums / Kit
• Bass
• Lead Vocal
• Backing / Harmony Vocals
• Guitars
• Keys / Synths
• FX (shared reverbs, delays, risers - anything feeding from multiple tracks)
• Everything Else (a catch-all for whatever does not fit elsewhere)

Not every song needs all eight - group by what is actually useful. A four piece rock song might only need Drums, Bass, Guitars, and Vocals.

The simplified four stem approach. For a lighter touch that still gives real flexibility, a well known alternative (associated with veteran mastering engineer Bob Katz) uses just four files:

1. Lead vocal only
2. Backing vocals only
3. Instrumental (everything except vocals)
4. Full mix (not a stem to be processed - just a reference of how it should all sound combined)

This gets you most of the benefit - especially the ability to rebalance the most important element, the lead vocal - without multiplying the number of files or adding much to the turnaround time or cost.

This is the part people get wrong most often.

Remove:

• Any limiter, maximizer, or clipper on your master bus - this exists purely to make things louder and takes away the mastering engineer's options.
• Any EQ, saturation, or "glue" compression that was only bolted on at the very end to make the mix sound bigger or brighter, rather than being part of the mix itself.

Keep:

• Any processing that is genuinely part of the sound - an EQ move, compressor, or saturation plugin that shapes the character of an instrument and was part of your creative decisions during mixing.
• All automation - volume rides, panning moves, effect sends. If it was a deliberate choice during your mix, it should carry through to the stems.

The test to apply: if you muted every plugin that is "just for level," would the song still sound like your mix, only without the final loudness push? If yes, you have drawn the line in the right place.

Clear naming saves real back-and-forth. A simple, readable convention works best:

ArtistName_SongTitle_StemName.wav

Example:
TheHollows_Fallout_Drums.wav
TheHollows_Fallout_Bass.wav
TheHollows_Fallout_LeadVocal.wav
TheHollows_Fallout_BackingVox.wav
TheHollows_Fallout_FullMix.wav

A few rules of thumb:

• Avoid generic names like "Audio 1" or "Track 5" - you would be surprised how often these show up.
• Stick to letters, numbers, underscores, and hyphens. Skip special characters that can behave unpredictably across operating systems.
• Keep everything for one song in a single folder, and zip the whole folder before uploading - it keeps files together and avoids anything going missing in transfer.

• Import all your exported stems into a fresh, empty session and play them back together at unity gain. They should sound identical to your original mix. If they do not, revisit your export.
• Solo each stem individually and check for clipping - no single file should peak above 0 dBFS.
• Quickly check your mix in mono. Phase issues between stereo elements (a common culprit: stereo-mic'd acoustic instruments) can cause parts to thin out or disappear entirely when summed, and it is much easier to catch this before sending than after.

• Leaving a limiter or maximizer on the master bus "just to hear it louder" while exporting stems.
• Stems that do not all start from the same point in the timeline.
• Trimming off reverb or delay tails at the end of a stem.
• Vague or generic file names that force the engineer to guess what is in each file.
• Sending only the stems with no reference mix, no reference track, and no notes - context saves time and gets better results.
• Being indecisive about vocal level and not communicating it - send alternates rather than leaving the engineer to guess.