Release-Ready Audio: Checks Before You Upload

A mix can sound finished in your studio and still cause trouble in the week before release. The file turns out to be an earlier bounce. The last second of reverb is cut off. A limiter pushed the peaks over zero and a store’s encoder made them worse. Or the track sounds right in the room and loses its bass on a phone. None of these is a mixing mistake in the artistic sense. They are delivery mistakes, cheap to catch before you upload and awkward to fix afterwards.

This guide is a checklist for the audio file itself: the format WBBT asks for, how to check clipping, silence, channel balance and the end of the file, what Spotify, Apple and YouTube have published about loudness, and how to judge a track on small speakers. It leaves out metadata and artwork, which have their own guide, and it does not teach mixing. We are a distributor, so we have an interest in releases passing review the first time, and we have kept the advice to what the sources say and what the product does. Every external page below was read on 25 October 2026.

What WBBT asks for

The WBBT FAQ gives the current rule for standard audio releases: WAV files at 16-bit or 24-bit, at 44.1 kHz or 48 kHz, made from the best source master you have. It also asks you to check four things before you upload: clipping, silence, channel balance and the actual end of the file. Dolby Atmos and other specialist deliveries need a separate specification from support.

The FAQ sets no loudness target, and neither does this guide. Loudness is a creative decision, and the stores treat it in their own ways, as the section on loudness explains. What review looks at is whether the file is usable: whether it is the right recording, whether it is clean, and whether it is what you meant to send. When a release is sent back for an audio problem, the reason appears in your account or in a message from support, and the FAQ’s advice is to fix that specific problem rather than upload the same files again.

Format: WAV, bit depth and sample rate

Three decisions matter, and all three are about not damaging the master.

• Keep the sample rate you finished at, if it is accepted. If the session ran at 44.1 kHz or 48 kHz, export at that rate. If you worked at 88.2 or 96 kHz, convert once, with a good sample-rate converter, to 48 kHz or 44.1 kHz, because those are the rates the upload specification names. Apple’s brief warns that downsampling can create aliasing, so do the conversion once and listen to the result instead of repeating it.
• Pick one bit depth and stay with it. 24-bit is fine and gives more room. If you reduce to 16-bit, add dither once, as the last step. Apple describes dither as a trade-off: it reduces the distortion of the reduction at the cost of a little added noise. Truncating without dither can cause quantisation distortion.
• Never upsample or pad. Apple’s brief says that upsampling does not recover or add information, and that you should not “bit-pad” 16-bit files into 24-bit. A 44.1 kHz file stretched to 96 kHz is not high resolution, and an MP3 saved as WAV is still an MP3.

How to spot a lossy file saved as WAV

If a file reached you from a collaborator or an old drive and you are unsure of its origin, open it in a spectrum view. Many lossy encodes remove everything above a cut-off somewhere in the upper range, and the picture shows a sharp ceiling with nothing above it. A free tool such as FFmpeg can draw one with its showspectrumpic filter. A ceiling is a warning, not proof, because some genuine masters have little high-frequency content, but it is a good reason to look for the original session export before you upload.

Clipping and headroom

Digital audio has a hard ceiling at 0 dBFS. A signal that tries to go beyond it is clipped, which sounds like distortion. Two separate problems hide here.

The first is plain sample clipping, where the file itself contains flat-topped peaks. Look at the waveform of the loudest section at full zoom. Runs of samples sitting at the same maximum value are clipping. Fix it at the source: lower the level before the limiter, or lower the limiter’s output ceiling, and export again.

The second is subtler. Playback chains often oversample the signal before converting it to sound, and the stores encode your file to a lossy format for many listeners. Apple’s brief explains that if the original data already sits at 0 dBFS, oversampling can produce clipping, and it recommends leaving at least 1 dB of headroom. It also notes that levels that show no overs in the PCM file can still clip once encoded, which is why Apple supplies a tool, afclip, to check an encoded file as well as the master. Spotify’s loudness page gives similar numbers in different words: keep the true peak below -1 dB TP, and below -2 dB TP if the master is louder than -14 LUFS integrated, to avoid extra distortion in the lossy formats.

The practical rule is to set the output ceiling of your final limiter to -1 dBTP or lower, and to use a true-peak meter, not only a sample-peak meter, to check it. Louder, more compressed masters need more margin, so -2 dBTP is a sensible ceiling for a dense master.

Loudness: what the stores say, and why there is no single target

Streaming services adjust playback level so that listeners do not have to reach for the volume control between tracks. That changes what “loud enough” means. The table gives what the pages we read say, and nothing more.

Service	What its own page says
Spotify	Tracks are adjusted to -14 dB LUFS according to the ITU 1770 standard. Premium listeners can choose Loud (-11), Normal (-14) or Quiet (-19). Louder masters are turned down, and softer ones are lifted only as far as the peak headroom allows. The web player and third-party devices such as TVs do not use normalisation. Spotify asks for one high-quality stereo master per track.
Apple Music	Sound Check is a playback setting that plays songs at the same volume. Apple’s brief says it measures each track, stores that information, and then raises or lowers playback accordingly, and that loud masters end up played back lower, which can make them sound weaker. The brief we read publishes no single loudness target.
YouTube	The help page describes Stable volume, a viewer setting that balances the range between quiet and loud parts. Creators cannot control it directly. The page states no target level, so we do not quote one.

Put together: there is no universal LUFS number, the rules differ by service and by listener setting, and a quoted “-14 for everything” is Spotify’s figure, not a rule of the industry. The sensible approach is to choose a loudness that suits the music, check that the peaks are clean, and accept that the services will turn loud masters down. Pushing a master harder gains nothing at Spotify’s default setting and costs dynamics. Measure your integrated loudness so that you know where you are, but do not chase a number.

A worked example, with invented readings

The single below, its readings and its fixes are invented to show how the checks fit together. They are not a real release and the numbers are not a standard.

Example Sound has bounced “Harbour Lights” and runs three commands in the free FFmpeg tool. The first, ffprobe, shows the file’s format. The second measures loudness and true peak. The third looks for silence.

ffprobe -hide_banner harbour-lights-master.wav
ffmpeg -hide_banner -nostats -i harbour-lights-master.wav -af ebur128=peak=true -f null -
ffmpeg -hide_banner -i harbour-lights-master.wav -af silencedetect=noise=-60dB:d=1 -f null -

FFmpeg’s documentation describes the ebur128 filter with a true-peak mode in its peak option, and the silencedetect filter with its noise and duration settings (FFmpeg filters documentation, accessed 25 October 2026). Any other loudness meter and spectrum view does the same job. The readings come back like this:

Check	Reading	Verdict	Action
Format	WAV, 24-bit, 48 kHz, stereo	Accepted	None
Integrated loudness	-9.6 LUFS	Very loud	Fine as a choice; expect it to be turned down
True peak	+0.4 dBTP	Over the ceiling	Set the limiter ceiling to -2 dBTP and export again
Silence at the start	0.3 seconds	Acceptable	None
End of file	Reverb tail cut at 3:40.9	Abrupt	Re-export with the full tail, or add a short fade
Mono sum	Lead synth drops sharply	Phase problem	Check the stereo widener and the polarity

The loudness line shows why chasing level is a poor trade. Spotify’s page says it adjusts tracks to -14 dB LUFS, so at the Normal setting a master at -9.6 LUFS is turned down by 4.4 dB (14 minus 9.6). The extra limiting that bought that loudness is paid for in dynamics, and the listener hears the track at the same level as a cleaner master. Example Sound re-exports with a lower ceiling, hears no difference in level on a service that normalises, and keeps more punch.

Silence, the start and the end of the file

Listen to the first and last ten seconds of every file with headphones, at a moderate level, as a last check.

• The start. A little silence before the first note is normal. A click, a breath of noise or a clipped first transient is not. Check that the track begins where you intend, and not a few frames late because a fade-in was applied twice.
• The end. Let the tail ring out. A reverb or a held note chopped by the export boundary is the commonest ending fault. If you want a fade, make it a deliberate fade, not a cut. Look at the waveform to confirm that the file ends where the audio does, with no long dead stretch after it.
• The middle. A silence detector flags unplanned gaps, such as a muted track that left a hole or a dropout from a bad bounce. It also flags gaps you meant to leave, so read the report rather than trust it.
• Digital silence. A long run of absolute zero at the end of a file can make a player seem to stop dead. A noise floor that fades out naturally is easier on the ear.

Channel balance, mono and phone speakers

A file can look correct and still be wrong on one side. Check these three things.

1. Balance. Play the track and confirm that both channels carry audio at a similar level. A stereo file with one side missing, or a mono file saved with everything in the left channel, is the kind of fault the channel balance check exists to catch.
2. Mono compatibility. Sum the file to mono. Anything that disappears or turns thin when you do is cancelling out of phase: a stereo widener, a doubled guitar with flipped polarity, a wide synth pad. Many phones, smart speakers and shop systems play mono or near-mono, so the lead elements should survive the sum.
3. Small-speaker translation. Phone and laptop speakers cannot reproduce the deepest bass, so a sub-bass-heavy track can seem to lose its low end and its weight. The fix is not to boost the bass until a phone moves air. Give the bass a clear upper harmonic, such as gentle saturation, so that the part carrying the notes is audible on a small speaker, and keep the true sub where it belongs. Check the track on a phone speaker, on cheap earbuds and in a car if you can, and compare it with a commercial release you know well. Apple’s brief recognises the same spread of listening: its examples run from AirPods on a loud train to a HomePod in a quiet room.

None of this means flattening the mix. It means knowing what will vanish and deciding whether you mind.

Is it the right file?

The most expensive audio mistake is also the dullest: uploading the wrong version. Bounces pile up in a session folder as “final”, “final2” and “final_master_NEW”, and the one you sent to the mastering engineer is not always the one you approved.

• Name approved masters with the track number, the title and the version, and keep earlier bounces in a separate folder.
• Play each file you are about to upload from start to end, once, in track order. It takes minutes and catches the wrong version, the swapped track and the loop that never ends.
• Check that clean and explicit versions are in the right slots, and that an instrumental has no vocal bleed.
• If you finished the track in a different session from the one you remember, check the file’s date and length against your notes.

In the Release Wizard you can save a draft and come back, which makes it easy to do this check once the files are in the draft and not only on your desktop. If you work with an assistant, one connected through WBBT MCP can upload audio to a draft and run the same checks as the wizard, but it cannot send the release for you, so you still listen to the files yourself.

Remasters, AI-assisted tracks and specialist formats

• Remasters and reissues. A new master of an older recording is still a new file to check. The guide to re-releasing old music explains how a reissue relates to the original’s identifiers. Apple adds a condition for its own Digital Masters badge: remastered content must begin with a high-resolution digitisation of the original analogue source, and sound clearly better than the earlier version. That is a stricter test than the upload specification, so do not claim the badge unless you have been through that process.
• AI-assisted audio. The audio checks above apply to every file, whatever made it. Generated tracks have rights and declaration questions of their own, covered in how to distribute AI music.
• Dolby Atmos and club releases. A spatial mix is a separate deliverable, not a renamed stereo file, and it needs a specification from support. See Dolby Atmos distribution and the electronic music release checklist.

A twelve-point check before you upload

1. Find the final export. Confirm that it is the approved version, not an earlier bounce.
2. Check the format. WAV, 16-bit or 24-bit, 44.1 kHz or 48 kHz, with no upsampling.
3. Look for a lossy origin. Open a spectrum view if the file’s history is unclear.
4. Look for clipping. Zoom into the loudest section and check for flat tops.
5. Check the true peak. Keep it at -1 dBTP or lower, and at -2 dBTP for dense masters.
6. Measure integrated loudness. Know the figure and make the choice deliberately.
7. Listen to the first ten seconds. No clicks, no noise burst, no late start.
8. Listen to the last ten seconds. The tail is whole, or the fade is intended.
9. Run a silence check. Review each reported gap.
10. Check both channels. Both sides carry audio at similar levels.
11. Sum to mono. The lead elements stay put.
12. Play it on a phone. Decide whether the missing bass matters.

Then check the rest of the release with the metadata and artwork requirements guide, which covers the cover, titles, credits and copyright lines, and read how ISRC and UPC codes work if the audio has been released before. The Free plan needs a release date at least 14 days ahead (see the pricing page), so leave time to fix a file if review returns it.

Why audio gets a release returned, and the fix

The FAQ lists audio problems among the common reasons for a return, alongside rights conflicts, metadata mismatches and unsuitable artwork. It does not break them down, so the table lists typical causes of an audio return based on the checks above. They are not figures from WBBT’s review data.

Typical problem	Likely cause	Fix
Distortion or clipping	Limiter ceiling at 0 dBFS or above	Lower the ceiling to -1 dBTP or less and export again
Low-quality source	A lossy file saved as WAV, or an upsampled file	Find the original session export and upload the lossless master
Wrong or incomplete file	An earlier bounce, or a tail cut at the end	Re-export the approved version in full
One-sided or missing channel	A routing mistake, or a mono file placed in one channel	Fix the routing and export a true stereo or dual-mono file
Unexpected silence or gaps	A muted track, a bad bounce or a long dead tail	Trim the dead air and check the middle of the track
Wrong format	A rate or depth outside the specification	Convert once to WAV at 44.1 or 48 kHz, 16 or 24-bit

If the reason is unclear, ask support. Send the release title, the UPC or ISRC if you have one, the track affected and what you changed, so that the answer can be specific.

If you find a fault after upload

Before you submit, nothing has been delivered, so correcting a file is a matter of re-exporting it and checking again. After a release is live it is not so simple. The FAQ’s advice is to ask support before replacing audio, because a materially different recording may need a new ISRC and a new submission, and an identifier should not be reused for a different performance just to keep the play counts. A genuine fix to the same recording is a conversation with support, not a quiet re-upload. That is the strongest reason for the twelve-point check: it costs minutes now and is awkward to do later.

Frequently asked questions

What loudness should I master to for Spotify and Apple Music?

There is no universal figure. Spotify names -14 LUFS integrated with a true peak below -1 dB. Apple describes Sound Check, and its brief for masters asks mainly for clean, unclipped files with at least 1 dB of headroom. WBBT sets no loudness target, so master for the music and keep the peaks clean.

Does a louder master sound better on streaming services?

Usually not. Spotify turns louder masters down, and Apple says that loud masters are played back at a lower volume with Sound Check, which can make them sound weaker. A heavily limited master gives up dynamics and gains no loudness over a cleaner one.

Can I upload an MP3 if I do not have a WAV?

Look for the original export first, because WBBT asks for the final lossless master. Converting an MP3 to WAV only changes the container. The lost detail does not come back, and a reviewer or a listener may hear it.

Is 24-bit better than 16-bit for the file I upload?

WBBT accepts both. Use the bit depth your master was finished at. If you must reduce a 24-bit file to 16-bit, apply dither once at the very end, and never pad a 16-bit file into 24-bit and describe it as high resolution.

What if I notice a mistake after the release is live?

Ask support before replacing audio. A different recording may need a new ISRC and a new release submission, and an identifier should not be reused for a materially different performance. See the WBBT FAQ.

Do I need a separate master for each store?

No. Spotify says one high-quality stereo master per track is enough, and the stores make their own encodes for playback. Specialist formats such as Dolby Atmos are separate deliverables and need their own specification from support.

Next steps and sources

Sources read for this article, all accessed on 25 October 2026:

• Loudness normalization on Spotify, Spotify Support for Artists.
• Apple Digital Masters, Apple, with its technology brief for masters (PDF).
• Adjust the volume in Music on Mac, Apple Support, for the Sound Check setting.
• Video and audio quality enhancements, YouTube Help, for Stable volume.
• FFmpeg filters documentation, for the ebur128, silencedetect and showspectrumpic filters.

Store policies change, so check the current pages before you rely on a figure. To go further, read the pillar guide to release metadata and artwork requirements, then how a release goes from files to live. The music distribution page describes the route to the stores, the pricing page compares plans, and the WBBT FAQ answers questions on audio, artwork and credits.