The final stereo master is the smallest part of the delivery. A placement normally needs the full mix, an instrumental, a version with the lead vocal removed but the backing vocals intact, an a cappella, at least one shortened edit, and a stem set that sums back to the master. All of it at 48 kHz and 24 bit, with true peak headroom left in, named so that a file makes sense on its own. Separately, the production fills in a cue sheet, and that document rather than the audio decides whether anyone gets paid.
Why does the final mix on its own stall an edit?
Because an editor is cutting picture, and picture moves.
A scene runs 47 seconds and your song runs 3:38. The dialogue starts nine seconds in. Somebody laughs over the second chorus. Every one of those is a problem that gets solved by having another file, and none of them gets solved by having a better master.
When only the master exists, the editor does the arithmetic and takes the cheaper route, which is to use a different track. That decision happens in an afternoon, without a conversation, and nobody tells you it happened. The whole point of preparing files in advance is to stay in the running during the hours when nobody can reach you.
What is a TV mix, and how is it different from an instrumental?
An instrumental has no vocals at all. A TV mix keeps the backing vocals and removes the lead.
That distinction gets confused constantly, and the two files do opposite jobs. An instrumental goes under dialogue, where any voice competes with the actors. A TV mix goes under a scene with no dialogue, where the record still needs to sound like a record. Strip the harmony stacks out and the arrangement usually collapses, because in most modern pop the stacks are carrying the chord.
Two more versions earn their place. An a cappella, which gets used more often for trailers and montage than anyone expects. And a version with the lead vocal down 6 dB rather than out, which is the one that solves the "dialogue for the first line, then the song opens up" problem in a single file.
How do you export stems that actually sum back to the master?
Print them through the master bus, not around it.
The common failure is exporting each group with the bus compressor and limiter bypassed, which gives a clean set of files that sound nothing like the record when you add them together. If the bus chain is doing 3 dB of gain reduction on the mix, that processing is part of the arrangement.
The workable method is to print each stem with the full master chain active, with every other group muted. Compression on the bus then responds to less material, which is why the sum will not be exact, and a difference under about 0.5 dB is usually accepted. If the bus compressor is doing more than 2 or 3 dB, print the stems dry and supply the master chain settings as a note instead, because at that point the sum will drift audibly.
Four to six stems is the useful count: drums, bass, music, lead vocal, backing vocals, and a sixth for whatever makes the record identifiable. On a Ukrainian arrangement that sixth stem is usually the sopilka or the telenka, because that is the part someone will ask to raise or lose.
Every stem starts at the same timecode, including the ones that are silent for the first minute. A stem that starts where its first note starts is the fastest way to have a set of files nobody can line up.
What sample rate, bit depth and level should the files be?
48 kHz, 24 bit, and enough headroom that nothing further down the chain has to make a decision for you.
48 kHz because video runs at 48 kHz, and at 24 frames per second it gives exactly 2000 samples per frame. 44.1 gives 1837.5, so a cut on a frame boundary does not land on a sample boundary, and the file gets resampled somewhere in the chain by whichever converter happens to be in the edit suite.
On level, the destinations disagree and they disagree by a useful amount. US broadcast works to ATSC A/85, which targets -24 LKFS with a true peak ceiling of -2 dBTP. European broadcast works to EBU R128 at -23 LUFS. Streaming sits lower again: the published Netflix sound mix specification asks for -27 LKFS, dialogue-gated, measured with ITU-R BS.1770-1 across the whole programme.
Read those three numbers together and the practical instruction is the opposite of what a release master needs. Do not deliver a song crushed to -8 LUFS integrated. It will be turned down by 15 to 19 dB, and everything the limiter did to make it feel loud arrives as flatness rather than as impact. A placement master with 2 to 3 dB more dynamic range than the streaming release is the version that survives.
How should files be named, and what should be inside them?
Name them so that one file, found alone on a drive in six months, still identifies itself.
A workable pattern is artist, title, version, then the technical spec: `Chilibi_SongTitle_TVMix_48k24b.wav`. No spaces, no accented characters, no "final_v3_NEW". The person opening it is not you, and their session is not organised the way yours is.
Inside the file, the Broadcast Wave Format bext chunk carries an origination date, a reference timestamp and a free description field that travels with the audio through most professional software. Filling it takes a minute on export and it survives being renamed by three different people. Above 4 GB, which a long stem set will pass, use RF64 rather than standard WAV.
Who fills in the cue sheet, and why does it decide the money?
The production does, usually the music department or the post supervisor, and it is worth understanding even though it is not your document to file.
A cue sheet is the schedule of every piece of music in a programme: title, writers, publishers, their shares and their society, plus how each cue is used and how long it runs. Performing rights organisations distribute broadcast royalties from it. ASCAP publishes what the fields mean, and reading their page once is a better use of twenty minutes than most things a writer does in a week.
What you can control is upstream of it. Your work needs to be registered with your society with the title spelled exactly as it will appear, and your splits need to be agreed in writing before the placement rather than after. When a cue sheet arrives with a misspelled title or a missing writer, the money goes into an unmatched pool, and getting it out afterwards is slow.
Frequently asked
Do I need Dolby Atmos versions? Not for a song placement. Atmos deliverables apply to the programme mix, which is the responsibility of the re-recording mixer, not of the song.
Should I deliver a :30 and a :60 edit? Yes, and make them musical rather than truncated. An edit that ends on a downbeat with a real ending gets used. One that fades at 30 seconds gets recut by an assistant who does not know the song.
What if the supervisor only asks for the master? Send the master, and mention in one line that instrumental, TV mix, a cappella and stems exist. That sentence costs nothing and answers the question they will have on the following Tuesday.
If you have a placement coming and no files ready
Preparing a delivery set is a session, not a favour, and it goes faster on a record while the mix is still open. If you have a song that might be pitched for picture, send the session and the target destination, broadcast or streaming, and I will tell you which versions are worth printing and which are a waste of a day.
Full credit list at /credits, background and contact at /about.