🎧 Listen to this article
On This Page
0%TL;DR — a release-ready audio pack is five files, not one:
- Master — the loudness-checked stereo file your distributor gets.
- Instrumental — sync placements, live sets, lyric videos.
- Clean edit — radio, editorial playlists, ad platforms.
- Acappella — remixers, fan edits, short-form hooks.
- Stems — collaborators, DJs, and every future re-cut.
Build all five on release day. Rebuilding them later costs a week.
Most indie releases ship exactly one file: the master. Then three weeks later a sync agent asks for an instrumental, a curator asks for a clean edit, and a remixer asks for the acappella — and the session is either on a dead drive or locked in a project file the collaborator can't open. The pack takes about an hour if you assemble it while the mix is still in front of you. Reconstructing it from a distributed MP3 six months on is a different job entirely.
This is a workflow post, not a mixing post. It assumes you have a finished mix and want the deliverables around it.
Do this, not that
| Do | Don't | Why it matters |
|---|---|---|
| Export the instrumental from the same master you're distributing | Re-mix a "no vocals" version months later | A separately mixed instrumental drifts in level and tone. Sync editors cut it against the vocal version and hear the seam. |
| Archive a lossless master (24-bit WAV) | Archive only the MP3 you uploaded | Every derivative inherits the artifacts of the file you started from. Lossy in, lossy forever. |
| Build the clean edit by editing — reverse, duplicate, or drop a syllable | Mute the word to silence | A hole in the vocal reads as a mistake. A crafted edit reads as an intentional radio version. |
| Check integrated loudness and cap true peak around −1 dBTP | Master as loud as the limiter allows | Streaming platforms normalize playback anyway. Overshoot costs dynamics and risks distortion after lossy transcode. |
| Render a 15–30 second hook the same day | Wait until a video editor asks on a deadline | The hook is the single most-reused asset you'll make. It gets requested at the worst possible moment. |
| Use one strict naming convention across the pack | final_v3_REAL.wav | Distributors, collaborators, and your own future self all parse filenames before they open anything. |
| Repair noise and hum on the source, before the mix | Run noise reduction across a finished master | Broadband processing on a completed mix smears cymbals and reverb tails. Fix it upstream in noise reduction. |
| Keep every stem starting at sample zero | Trim leading silence per stem | Stems that don't share a start point are useless the moment somebody imports them into a different session. |
What files does a music release actually need?
| Deliverable | Who asks for it | Practical format |
|---|---|---|
| Master | Distributor, DSPs | 24-bit WAV, source sample rate |
| Instrumental | Sync agents, lyric video, live playback | Same as master |
| Clean edit | Radio, editorial playlists, ad platforms | Same as master |
| Acappella | Remixers, short-form editors | Same as master |
| Stems (4–6) | Collaborators, DJs, remix packs | WAV bundle, all sample-aligned |
| Short-form cutdowns | TikTok, Reels, Shorts, ads | 15–60s, one hook per file |
| Metadata sheet | Distributor, PRO, sync library | Plain text or CSV |
Seven items, one hour. Here's the reasoning behind each.
The master is an archive decision, not just an upload
Whatever you hand your distributor is the file the world hears, but the file you keep is the one every future deliverable is cut from. Keep the highest-resolution bounce you have, uncompressed, with a checksum-friendly name. If a supervisor asks for a version at a different length in eighteen months, the difference between a lossless archive and a downloaded stream is the difference between an edit and a re-record.
The instrumental is the highest-return file you'll make
Instrumentals get placed. They back podcast intros, YouTube segments, brand spots, and live sets where the vocalist isn't in the room. They're also what lyric videos and karaoke cuts are built from. If you still have the session, bounce it with the vocal bus muted and nothing else changed. If you don't, separate it from the master with stem splitting — modern separation on a well-mixed track holds up well for backing use, and the free path through the vocal remover is enough to test whether the result is usable before you commit to anything.
The clean edit is an editing job, not a mute
Playlist editors and ad platforms reject explicit audio quietly — you usually find out by not being added. Make the clean version by editing at the syllable level: reverse the word, duplicate a neighboring phoneme, or drop it under a fill. Silence draws attention to itself; a competent edit doesn't. Do this on the vocal alone if you have stems, then re-render, so the instrumental bed stays continuous underneath. Sample-accurate edits are straightforward in a timeline like the browser-based DAW, where you can nudge a region and audition against the original.
The acappella is what makes other people promote you
Remixers, mashup editors, and short-form creators need isolated vocals. Handing them a clean acappella is the cheapest distribution you will ever buy — someone else does the work of putting your voice in front of an audience you don't have. Publish it with a clear license note. If your archive is only the stereo master, vocal isolation gets you there; check the result on headphones for artifacts in dense passages before you release it.
Stems are insurance against your own future
Four to six stems — drums, bass, vocals, and whatever carries the harmony — cover almost every request that arrives after release: a live backing track missing one part, a remix pack, a re-cut for a different edit length. Export them all from the same start point at the same level, so anyone can drop the folder into any session and hit play. Producers who need finer control can separate further; the premium catalog on Pro covers 45 instruments, which is a menu of what you can isolate, not a promise that every song contains 45 usable parts. Tools like LALAL.AI and Moises solve the same problem with different pricing models — worth comparing if separation is the only thing you need.
Short-form cutdowns decide whether anyone hears the song
Pick the hook, not the intro. Export a 15-second and a 30-second version, both starting on the downbeat of the strongest phrase. Vertical-video platforms normalize loudness on their own, so don't crush the cutdown harder than the master — it just sounds smaller after normalization. If you need an alternate bed, a track edit, or a transition sting for the video, generating one in music studio is faster than licensing a stock cue.
Metadata is the deliverable people forget
ISRC, writer splits, publisher, BPM, key, explicit flag, and release date belong in one plain-text file that ships with the audio. Sync libraries and distributors will each ask for the same information in a different form; having it in one place turns a twenty-minute lookup into a copy-paste.
How do you build the pack if you never kept the stems?
This is the common case — the mix was done on someone else's machine, or the drive is gone, and all that survives is the released track. The recovery order that works:
- Start from the highest-quality file you can find. A distributor's original upload beats a streamed copy.
- Separate vocals from instrumental first, and listen to both in isolation before doing anything else. If the instrumental has vocal bleed in the chorus, the rest of the pack inherits it.
- Split the instrumental into drums, bass, and the remaining parts only if you actually need them. More stems means more chances for artifacts.
- Rebuild the clean edit from the recovered vocal, not the full mix — you get a real edit instead of a hole punched through the whole arrangement.
- Cut the short-form hooks last, from the finished derivatives, so every asset in the pack matches.
One pass through separation gets you an instrumental, an acappella, and a stem set from a single stereo file. That's four of the seven deliverables recovered from a track you thought was closed.
Release-day checklist
- Lossless master archived with a dated filename
- Integrated loudness checked, true peak capped
- Instrumental bounced or separated, listened through once end to end
- Clean edit made by editing, not muting
- Acappella exported and artifact-checked on headphones
- 4–6 stems, all starting at sample zero
- 15s and 30s hooks rendered
- Metadata sheet written
- Whole folder backed up somewhere that isn't your laptop
FAQ
Do I need stems if I have no plans to remix?
Yes — the plans arrive later, from other people. A backing-track request for a live show, a brand asking for a shorter cut, a collaborator wanting to re-sing one line: all of them are trivial with stems and expensive without them. Ten minutes on release day, or a rebuild months later.
Can separated stems be released commercially?
Rights-wise, they're derivatives of your own recording, so if you own or control the master you control the stems. The practical question is quality: separation is excellent for backing tracks, DJ use, and short-form, and it is auditioned rather than assumed for a flagship release. Listen to solos before you publish.
What loudness should I master to?
Published normalization targets on the major streaming services sit in roughly the −14 to −16 LUFS range, and they turn loud masters down rather than rewarding them. Cap true peak around −1 dBTP to leave headroom for lossy encoding. Platform-by-platform detail is its own topic — see the loudness guide on the blog.
How is this different from what Suno-style generators produce?
Generated tracks usually arrive as a finished stereo file, which means they have the same missing-deliverables problem as a mix you lost the session for. The recovery workflow above applies identically: separate first, then build the instrumental, acappella, and cutdowns from what you get.
How long does AudioPod keep my generated files?
Retention is tier-based: 1 year on Basic, 2 years on Creator, 3 years on Pro, 5 years on Studio, and indefinite on Enterprise. Downloading or reopening a file inside the last 30 days extends its window, and anything you've marked featured, public, or shared is never removed automatically. Job history stays in the dashboard either way, so a project can be re-rendered in one click. Details are on the pricing page.
What does the whole pack cost to produce?
Separation and generation are metered from your monthly credits — the free Basic tier is enough to run a full pack for one single and decide whether the output holds up. Creator at $20/mo covers a regular release schedule; Pro at $50/mo adds the 45-instrument premium separation catalog for producers who need finer splits.
The point
A release is not a file, it's a set. The artists who get sync placements, remixes, and playlist adds are usually not the ones with better songs than everyone around them — they're the ones who can answer "can you send me the instrumental?" the same afternoon. Build the pack while the mix is still open, name it properly, and put it somewhere you'll find it in two years.
