The question nobody can answer by ear
You have been working on the same track for weeks. You know every transition, every breath before the second chorus, every decision you argued with yourself about. And that is exactly the problem: after twenty passes your ears have stopped being a reliable instrument. Everything sounds normal because everything sounds familiar.
"Is it ready?" is usually treated as a feeling. It is not. A good part of the answer is measurable, and the measurable part is the one that stops a release from sounding amateur. Six numbers decide whether your file will survive being encoded, normalised and played back on a phone speaker. Get them right and what remains is genuinely an artistic decision. Get them wrong and no amount of taste will rescue the result.
This guide walks through those six checks, with the actual target values and, just as importantly, the order to fix them in.
1. Loudness: are you at -14 LUFS?
Integrated LUFS measures the perceived loudness of your whole track, not its peak level. It is the number streaming platforms use to decide how much to turn you up or down. Spotify, Apple Music and YouTube all normalise to roughly -14 LUFS.
This changes the maths completely. Master at -8 LUFS and the platform simply turns you down to -14, so you gave away your dynamics for nothing: the loudness is gone, the compression stays. Master at -20 LUFS and you are relying on the platform to turn you up, which does not always happen, so your track can feel thin and distant next to everything around it in a playlist.
Concrete action: measure integrated LUFS over the entire track, not over the loudest chorus. If you are between -15 and -13, you are in the right zone. Outside that window, adjust the master gain before touching anything else in the chain.
2. True peak: the -1 dBTP ceiling
Your DAW shows sample peaks: the height of the individual digital samples. True peak is different. It estimates the height of the analogue waveform that will be reconstructed between those samples, and it is almost always higher.
This matters because streaming platforms re-encode your upload into a lossy format such as AAC or Ogg Vorbis. That process introduces intersample peaks. A master sitting at 0 dBTP looks perfectly clean in your project and then distorts after encoding, on a file you never got to hear before your listeners did.
Concrete action: set your final limiter ceiling to -1 dBTP. Never go above -0.3 dBTP under any circumstances. This is the very last thing you touch in the chain, because any later change to EQ, compression or gain invalidates the ceiling you just set.
3. Clipping: damage that survives every export
Clipping happens when samples are pushed past full scale and the tops of the waveform are cut off flat. Unlike loudness, this is not a setting you can adjust later: it is destroyed information. Once the peaks are squared off, no amount of mastering brings the shape back. Export it, re-import it, master it again: the flat tops are still there.
Worse, it hides. A single clipped transient in a busy chorus is easy to miss on laptop speakers, and it turns into audible crackle on a good pair of headphones or a car system.
Concrete action: check whether any samples are pinned at full scale, and where. If there are, the fix is never on the master bus: go back and correct the gain staging at the source, then rebuild the chain from there. This is the first thing to repair, because every other correction downstream is built on top of it.
4. Loudness range: squashed or breathing?
Loudness range (LRA), measured in LU, describes how much the perceived loudness varies across the track. It is the difference between a song that breathes and one that sits at a single flat level from the first bar to the last.
Useful reference points:
- Below 3 LU: squashed flat. Technically loud, exhausting after ninety seconds.
- 3 to 6 LU: heavily compressed. Normal territory for modern pop, hip-hop and EDM.
- 6 to 12 LU: comfortably dynamic. Most well-mastered commercial releases live here.
- Above 12 LU: very dynamic. Typical of jazz, acoustic and orchestral material.
There is no single correct number. A ballad at 3 LU has lost its emotional shape; a club track at 14 LU will disappear in the mix on a big system. What matters is whether the range fits the genre and the song.
Concrete action: if your LRA is under 3 LU, ease off the master compression before doing anything else. You are not gaining loudness at that point, because normalisation takes it back anyway. You are only losing impact.
5. Phase and mono compatibility
Phase correlation runs from -1 to +1. At +1 the two channels are identical, effectively mono. Around 0 the image is wide. Below 0 the channels start cancelling each other when they are summed.
That last case is the dangerous one, because a great many listening situations sum to mono or something close to it: phone speakers, Bluetooth speakers, club PA systems, most laptops. A mix that sounds impressively wide on headphones can lose its bass entirely the moment it is summed.
Concrete action: keep mean phase correlation positive. Below roughly 0.2 the track is very wide and starts carrying real mono risk; below 0 it is genuinely out of phase and needs fixing, not tweaking. Also check the lowest value the track reaches, not just the average: one section that dips negative is enough to break a chorus on a phone.
6. Spectral balance against your genre
The last check is not an absolute value but a comparison. Every genre has a characteristic frequency balance, and a mix that would be perfectly judged for an acoustic record may be far too gentle in the low end for a hip-hop release.
What matters here is the shape of the balance across bands, not the absolute level. Two masters with identical band relationships but different volumes are the same mix, judged the same way. Comparing your spectrum against reference targets for your genre family tells you whether the low end is bloated, whether the midrange is scooped, whether the top is harsh.
Concrete action: compare your spectrum against reference tracks in your genre at matched loudness, so you are hearing the balance and not the volume difference. Look for consistent excesses across whole bands rather than individual peaks.
The order you fix things in
This is where most self-released tracks go wrong, and it costs hours. The checks above are a list, but corrections are a sequence, and doing them out of order means redoing work.
- Clipping and gain staging. Everything downstream is built on this. Fix it first or you are just making a damaged file louder.
- Corrective EQ. Remove build-ups and fill dips before any dynamics processing sees the signal.
- Dynamics. Set master compression once the spectrum is already balanced, otherwise the compressor reacts to problems you are about to remove.
- Phase and stereo image. Fix polarity, alignment and excessive width. This changes the summed level, so it has to come before loudness.
- Loudness. Now set integrated LUFS, with the chain finally stable.
- True peak ceiling. Absolutely last. Any earlier change would invalidate it.
Notice that loudness, the thing most people reach for first, sits in fifth place. Setting it early means setting it again after every subsequent fix.
Export settings to deliver
Everything above is judged on the file you render, so render it properly. Most technical faults in independent releases are not mixing mistakes at all: they are delivery mistakes.
- Format: WAV or AIFF, never MP3. Your distributor produces the lossy versions, and sending an MP3 means encoding an already lossy file a second time.
- Bit depth: 24-bit. There is no reason to deliver 16-bit unless a specific target demands it.
- Sample rate: the one you worked at, typically 44.1 or 48 kHz. Do not resample at export.
- Dither: only when reducing bit depth. Exporting 24-bit from a 24 or 32-bit session needs none.
- Tails: leave reverb and delay tails intact. Cutting them at the last bar is a common amateur giveaway.
- Session hygiene: no solo or mute left active, no bypassed plugin you meant to re-enable, no reference track in the export path.
Concrete action: after rendering, import the file back into an empty session and listen to the first and last ten seconds. It takes a minute and catches truncated tails, wrong fades and silence at the head.
Matching the delivery target
One master at around -14 LUFS with a -1 dBTP ceiling works acceptably everywhere, and if you only make one, make that one. Separate versions are an optimisation, not a requirement.
- Streaming: about -14 LUFS, ceiling -1 dBTP, dynamic range preserved. This is the file your distributor gets.
- Social media: roughly -13 LUFS, ceiling -0.5 dBTP, slightly more compression, a small presence lift around 2 to 4 kHz and reduced content below 40 Hz.
- DJ and club use: -8 to -11 LUFS as a separate file, which should never go to streaming platforms.
Concrete action: name the files so the versions cannot be confused at three in the morning. track_streaming_-14LUFS.wav beats track_master_FINAL_v3_real.wav every time.
The go or no-go pass
The last pass is deliberately boring, and it exists because everything above changed the file. Re-measure the final render, then listen once end to end.
- All six values inside their target ranges, measured on the final file rather than on the session.
- One full listen on headphones at high volume, hunting for clicks, pops and distortion rather than judging the music.
- One listen in mono to confirm nothing collapses.
- One listen on a phone speaker, which is where much of your audience will meet the track first.
- Filename, track title, artist name and featured credits spelled exactly as they should appear in the stores.
If all five pass, the file can leave your computer. If one fails, fixing it now is far cheaper than withdrawing a live release and restarting the distribution cycle.
What release-ready actually looks like
"Release-ready" is not a compliment, it is a state you can verify. It means loudness, peaks, dynamics, spectral balance and stereo image all sit within the targets for your genre, with no corrective action left outstanding. At that point the technical side is clean and any remaining decision is an artistic one.
This is precisely the kind of check that does not need an opinion. TotalAnalyzer measures these values directly from your file using ffmpeg, compares them against per-genre reference targets, and returns a Mix & Mastering To-Do List: the actions to take, already sorted into the correct processing order shown above. If nothing needs fixing, it tells you that too, which is the answer you actually wanted.
These measurements are objective and repeatable. The same file always produces the same numbers, because nothing is being estimated: loudness, true peak, clipping, loudness range and phase correlation are computed from the audio itself, not judged by a model.