One WAV per track, every file starting at bar 1 of the session, at the sample rate the song was recorded at and 24-bit, with inserts bypassed; a second short set of the tracks whose plug-ins define the sound; a rough mix that peaks around minus 6 dBFS; and a one-page text file with tempo, key, track count and what was printed. That folder takes about an hour to build for a 60-track session, in the order below.
Why does every file have to start at bar 1?
Because it removes the only alignment method that fails silently. A track exported from its first note carries its position only in the file's metadata. Broadcast Wave files store a TimeReference field, the count of samples since midnight, and a receiving DAW can read that field and place the file where it belongs (EBU Tech 3285). Whether it does depends on the import setting and the person, and a background vocal that lands 2 bars early looks correct until somebody plays it.
A file that starts at bar 1 has nothing to read. It gets dropped at zero and it is right. In Pro Tools that means selecting from the session start to the end of the last clip on every track and running Edit > Consolidate Clip, which renders the selection into one continuous file with the fades baked in (Avid). Fades and clip gain go into that file; inserts and automation do not. In Logic the same thing is File > Export > All Tracks as Audio Files, with the Range set to extend silence to the project end so that every file also ends at the same sample (Apple). Equal length matters as much as equal start: when all 60 files are 3:52.000 long, a missing region shows as a shorter file before anybody listens.
I leave 2 bars of pre-roll before the first downbeat. A song that starts on the one gets pushed to bar 3, so the count-in and any pickup exist in the files.
Dry, processed, or both?
Both, in two named passes. The first pass is the source with every insert bypassed: in Logic that is the Bypass Effect Plug-ins box in the export dialog, with Include Audio Tail on and Normalize off (Apple). Volume and pan automation stay out of this pass. The mixer sets levels; a vocal that arrives with rides printed into it cannot be un-ridden.
The second pass is smaller. Any plug-in that is part of the sound gets printed as its own file: the amp simulator on a DI guitar, the vocal chain with the tuning and the doubler, the sidechained pad, the synth patch with its own delay. These get the suffix _FX and sit next to the dry version, 01_LeadVox_DRY.wav and 01_LeadVox_FX.wav. The mixer can use either, and cannot rebuild a patch from a plug-in they do not own.
Master bus processing never leaves the building. A limiter on the stereo out that prints into 60 tracks is 60 tracks of a limiter the mixer will fight for a week.
What sample rate and bit depth, and when do I convert?
The session rate, 24-bit, no conversion on the way out. A song tracked at 48 kHz goes out at 48 kHz; one tracked at 44.1 stays at 44.1. The one conversion I do is on imports: a 44.1 sample dropped into a 48 session gets rendered to 48 inside the session, so that every file in the export folder reports the same rate. The P&E Wing's delivery recommendations for recorded music projects, revised in 2025, say the same thing about originals in stronger language: keep them at the resolution they were made at, and document it (Recording Academy).
How should the tracks be named and ordered?
By a number that fixes the order and a name that says what the file is. My pattern is a two-digit group number, an underscore, the source, and a suffix when there are two versions:
01_Kick_In, 01_Kick_Out, 01_Snare_Top, 01_Snare_Bot, 01_HH, 01_OH_L, 01_OH_R, 01_Room_L, 01_Room_R; 02_Bass_DI, 02_Bass_Amp; 03_Gtr_Rhythm_L, 03_Gtr_Rhythm_R; 04_Keys; 05_LeadVox_DRY, 05_LeadVox_FX; 06_BV_1 to 06_BV_8; 07_FX_Riser; 08_Click.
Stereo pairs go out as two mono files, L and R, unless the source was a stereo instrument. The name has to survive being read by someone who was not in the room.
What has to be finished before the export?
The edit. A mix engineer's work starts at the balance, and a comp that arrives half done comes back as a question. My order: comp the lead, then tune it, then time the doubles and backing vocals to it, then check every crossfade at the sample level, then export. The comp method is in comping a lead in two passes and the tuning settings in how much pitch correction a lead needs; both happen before this folder exists. If the drums were quantised, the consolidated files carry the quantised version and the raw pass goes in a subfolder called RAW.
Two things are deliberately left undone. Levels: every fader at unity, every pan at centre, except the pans that are part of the arrangement, a hard-panned double for example, which I note in the text file. And mutes: a region muted in the session is exported anyway, in its own file, with _ALT in the name.
What goes in the rough mix and the note?
The rough is the balance the artist has approved, exported from the same session at the same rate, peaking around minus 6 dBFS, with no limiter. It goes in the root of the folder, named with the song and the date. References go next to it, level matched, with a line on what each one is for; the method is in reference tracks that work.
The note is one page, plain text, and always the same fields: song title and artist; tempo and any tempo changes with bar numbers; key; time signature changes; session sample rate and bit depth; number of files; which tracks have an _FX version and what the plug-in was; intentional pans; anything the artist has said about the vocal sound; a lyric sheet with section names and bar numbers; the deadline; and a phone number for the person who can answer questions today. A tempo map exported as a MIDI file goes in the same folder when the song has more than one tempo.
How do I check the folder before it goes?
By importing it into an empty session at the session rate. Every file drops at zero. I press play, watch for a file that ends early, solo the click against the drums, and listen to the sum against the rough at matched level. If the sum through unity faders sounds like a loud version of the rough, the export is complete. If a part is missing, it is missing here rather than in the mixer's inbox.
The folder goes as one zip named with the song and the date, and the note is pasted into the message as well, so it exists even if the download does not. Records I have mixed are listed with sources at /credits.
Frequently asked
Should I send stems or individual tracks for mixing? Individual tracks, one file per source, all starting at bar 1. Stems are summed groups and belong to delivery after the mix, when the balance is decided. A mix engineer who receives a drum stem cannot move the snare.
Do I remove plug-ins before exporting tracks for a mix? Export twice. One pass with every insert bypassed, so the mixer gets the raw source, and a second pass of only the tracks whose plug-ins define the sound, named so the two are never confused. Never export the master bus processing.
What sample rate should mix files be? The rate the session was recorded at, 24-bit, with no conversion on the way out. A song recorded at 48 kHz leaves at 48 kHz. Converting to 96 kHz adds nothing and doubles the download.
If a session is about to leave for mix
Send the note first. Tempo, key, track count, sample rate and what was printed tell me in two minutes whether the folder is ready. Contact at /about.