How to master a song

This covers mastering a finished stereo mix for a streaming release, with loudness numbers taken from each platform’s own help pages, checked Sep 24, 2026. Short version: aim for a clean, balanced master, not the loudest one.

How to master a song: the short version

To master a song, you run the finished stereo mix through five steps: prepare the file with headroom, shape the overall tone with EQ, control dynamics with gentle compression, raise the level with a limiter, and check loudness and true peak against the platform’s numbers. Mastering isn’t mixing. Mixing balances the individual tracks, and mastering treats the whole stereo file at once. You need a finished mix, a DAW with stock EQ, compressor and limiter plugins, and a loudness meter that reads LUFS. For Spotify, the target is published: Spotify for Artists says to master at -14 LUFS integrated with true peak below -1 dB. We found no published LUFS target from Apple Music, YouTube or Amazon Music (checked Sep 24, 2026).

Before you open a single plugin, get these together:

• The mix, exported on its own. Not a bounce with a limiter slapped on it. More on that below.
• A loudness meter. Your DAW may have one built in. If not, free meter plugins show integrated LUFS and true peak, and that’s all you need.
• Two reference songs. Released tracks in your genre that sound the way you want yours to sound.
• Two ways to listen. Headphones plus at least one set of speakers, even laptop speakers. A third (phone, car) comes in at the checking stage.

Each step below is a section of its own. None of it needs gear you don’t already have.

Before you touch anything, it helps to know what mastering can do and what it can’t.

What mastering does, and what it can’t fix

Mastering is the last processing step before release: it sets the overall tone, loudness and consistency of a finished stereo mix so it sounds right on phones, earbuds, car speakers and big systems. It works on the whole song at once, so it can’t turn up a vocal without turning up everything else in the same frequency range. A master can brighten a dull mix, tame harsh highs, even out levels across the songs on an EP and bring a song up to a sensible loudness. It can’t fix a buried vocal, a kick and bass fighting each other, a bad edit or a noisy recording. Problems like those go back to the mix.

Apple’s mastering brief for Apple Digital Masters describes the job well: EQ, compression and limiting are “all different ways of controlling aspects of volume,” and mastering is about decisions on gain levels, dynamic range and frequency response. That’s it. No magic.

Here’s how to tell which stage a problem belongs to:

The honest bit: most “my master sounds bad” problems are mix problems. If you catch yourself boosting 3 kHz by 6 dB to find the vocal, stop and reopen the mix session. You’ll get a better result in five minutes there than in an hour of mastering plugins.

Since the mix does most of the work, the file you hand to the mastering stage matters a lot.

Prepare your mix for mastering

Prepare a mix for mastering in four steps: take the limiter and any loudness plugins off the mix bus, leave some peak headroom, export a full-resolution stereo WAV at the session’s sample rate, and pick two reference songs in your genre. Headroom means the loudest peak sits below 0 dBFS, leaving room for the mastering limiter to do its job. There’s no single required number, and a few dB below zero is plenty. Export at the bit depth you mixed in (24-bit or 32-bit float) and don’t upsample. Apple’s mastering brief asks for 24-bit sources and says upsampling doesn’t add any information. Keep a copy of the unmastered mix.

Why the limiter comes off. A limiter on the mix bus makes the mix feel finished, which is exactly why people leave it on. But then the mastering limiter is squashing something that’s already squashed, and you lose the punch you liked. If the mix-bus limiter is part of the sound, export two versions (with and without) and compare them at the mastering stage.

Export settings. Use the session’s own sample rate, whether that’s 44.1 kHz or 48 kHz. Apple’s brief says not to upsample and not to pad 16-bit files out to 24-bit, because neither adds anything real. Name the file so future you knows what it is: songtitle_mix_v4_24bit_48k.wav beats final_FINAL2.wav.

The edges of the file. Trim the silence at the start so the song begins where it should, but leave the tail of the last note and any reverb to ring out. Fades are easier to judge after mastering, since the limiter changes how loud the ending feels.

Reference songs. Pick released songs you’d happily be playlisted next to. Import them into your mastering session on their own track, so you can flip between yours and theirs without changing anything else.

With a clean file ready, you can build the chain.

The mastering chain, step by step

A basic mastering chain has four stages in this order: EQ, compression, optional color (saturation or stereo width), and a limiter last. EQ fixes the overall tone with small, broad moves, usually a dB or two. Compression glues the mix together with a low ratio and only a little gain reduction. Color is optional and often skipped. The limiter goes last because it sets the final level and catches peaks, with its output ceiling set below 0 dB so the file doesn’t clip once it’s encoded. Every DAW’s stock plugins can handle all four stages. Logic Pro also has a built-in Mastering Assistant that sets up a starting chain for you.

1. EQ. Flip between your mix and a reference, and listen for one thing at a time. Too much low end? Too little air on top? Harsh in the upper mids? Make wide, gentle moves. If a fix needs more than about 3 dB, it’s probably a mix problem.

2. Compression. Start with a low ratio (around 1.5:1 or 2:1) and set the threshold so the meter shows a dB or two of gain reduction on the loudest parts. A slow attack lets drum hits through. You’re aiming for the song to feel a bit more solid, not for an obvious pumping sound. If you can hear the compressor working, back it off.

3. Color (optional). Saturation adds warmth and harmonics; stereo wideners spread the sides. Both are easy to overdo, and wideners can make a mix sound thin on a phone speaker, which is mono. If you’re not sure it helps, leave it out.

4. Limiter. Set the output ceiling first, then raise the input until your loudness meter reads the level you’re after. Watch the gain-reduction meter. A few dB on the loudest hits is normal. If it’s clamping down all the time, the song will start to sound flat and tired, and turning it up past that point only makes it worse.

Dither. Only needed if you export at a lower bit depth than you worked in, like 16-bit for a CD. Apple’s brief describes dither as a tradeoff: it reduces distortion when cutting bit depth, at the cost of a little added noise. For a 24-bit streaming master, you can skip it.

A stock shortcut in Logic Pro. If you use Logic Pro, its Mastering Assistant analyzes the mix and builds a starting chain. Apple’s help page says it limits the output to -1 dBFS true peak, and that the loudness knob at its center position typically lands around -14 LUFS integrated. Treat it as a starting point you adjust, the same way you’d treat a preset.

The limiter is where you choose how loud the song ends up, which raises the obvious question.

How loud to master for Spotify, Apple Music and YouTube

Of Spotify, Apple Music, YouTube and Amazon Music, only Spotify publishes a loudness target for artists: its Spotify for Artists help page says to master at -14 LUFS integrated and keep true peak below -1 dB, or below -2 dB if your master is louder than -14 LUFS. Spotify turns louder songs down and quieter songs up to -14 LUFS on its default setting. Apple Music’s Sound Check levels songs in a similar way, and Apple’s mastering brief gives no LUFS target but recommends at least 1 dB of headroom. We found no music loudness target in YouTube’s help pages, and Amazon Music only describes a listener setting. A louder master doesn’t play louder on Spotify. It gets turned down.

Here’s what each platform’s own documentation says, as of Sep 24, 2026:

More detail on Spotify, since it’s the one with numbers. The loudness normalization page says Premium listeners can pick Loud (-11 LUFS), Normal (-14 LUFS) or Quiet (-19 LUFS). On Loud, Spotify applies a limiter to stop distortion. It normalizes a whole album at once, so the level differences between your tracks stay as you set them. And the web player and third-party devices like TVs and speakers don’t normalize at all, so some listeners hear your master exactly as loud as you made it.

Why true peak matters. Streaming services play lossy files, and encoding can push peaks higher than they were in your WAV. Spotify says it leaves 1 dB of headroom for lossy encodings, and Apple’s brief recommends at least 1 dB of headroom to avoid clipping on playback. That’s where the -1 dB ceiling on your limiter comes from.

What this means for you. You don’t have to hit -14 LUFS exactly. A master at -10 LUFS just gets turned down about 4 dB on Spotify’s default setting, and if you squashed it to get there, it now sounds squashed and no louder than everything else. Apple’s brief says it plainly: songs mastered loud get played back at a lower volume, “which can make tracks actually sound weaker.” The worry about sounding quiet next to other songs in a playlist is mostly solved for you by normalization. Pick the loudness where your song sounds best, and let Spotify handle the rest.

Once you’ve picked a level, the last job is making sure the file is actually right.

Check your master before you upload

Check a master in five passes before you upload it: meter it for integrated LUFS and true peak, level-match it against your reference songs, listen on at least three different systems, check the encoded file for clipping, and play it top to bottom once more with fresh ears. Level-matching matters most, because a louder version almost always sounds better for the first few seconds, which tricks you into pushing the limiter harder. Test where your listeners actually are: cheap earbuds, a phone speaker, a car. Fix anything you hear, in the master or, for a balance problem, back in the mix, then export again.

Level-match honestly. Turn your reference down (not your master up) until the two read about the same on the loudness meter, then compare. This is the same thing Spotify does to both songs anyway, so it’s the fairest test of whether yours holds up.

Listen on three systems. Each one tells you something different:

• Phone speaker: can you still hear the vocal and the hook? Is the bass completely gone?
• Earbuds: is anything harsh or tiring after a full play?
• Car or a bigger speaker: does the low end boom or rattle?

Check the encoded file for clipping. A WAV can look clean and still clip once it’s turned into a streaming file. Apple’s brief says levels that don’t show overs in the PCM file can still clip when encoded, and it describes a free Mac command-line tool, afclip, that checks for both on-sample and inter-sample clipping. On Windows, a true peak meter set to your limiter’s ceiling is the practical stand-in.

Give it a day. The last full listen works best the next morning. Ears get used to a sound fast, and what you thought was a “warm” low end at 2 a.m. can turn out to be mud.

The fear most people have here is putting out a song that’s clipped or dull and being stuck with it. That’s why this pass exists. And if you do find a problem after release, a corrected master of the same mix is usually treated as the same recording: IFPI’s ISRC FAQ says a remaster can generally keep the original ISRC. How that works with your codes is in what is an ISRC code.

If the checks keep turning up things you can’t fix, that’s a sign to try a different route.

Master it yourself, use an online service, or hire an engineer

TheDebutRoom’s view is that most independent artists should master their own demos and early singles, use an automated mastering service when they need a quick and consistent result, and pay a mastering engineer for the songs they’re putting real promotion money behind. Learning the basics costs nothing but time, and it makes you a better mixer, because you start hearing what a master can’t fix. Automated services like LANDR and Mixea, DistroKid’s mastering service, analyze your mix and apply processing for you in minutes, with a few controls to adjust. A human engineer brings trained ears, a treated room and an honest second opinion on the mix itself.

Which one fits depends on your own numbers:

• You release often and on a small budget. Master it yourself, or use an automated service for consistency across a lot of singles. Spend the money on the release, not the polish.
• You can’t tell what’s wrong, only that something is. Try an automated master and compare it level-matched to your own. If the automated one wins, listen closely to what it changed. That’s a free lesson.
• This song is getting a real push (ads, a pitch to playlists, a video). Pay an engineer. Send them the mix without the mix-bus limiter, your references and a note on what you like and don’t like. Ask whether they’ll flag mix problems before they master.

On the automated side, LANDR’s mastering and Mixea both take an uploaded mix and return a master. The details (price, file formats, how much control you get, whether it’s bundled with distribution) differ, and they’re covered in their own reviews: the LANDR mastering review and the DistroKid mastering review. Whichever you use, run the result through the same checks as your own master. An automated master is a starting point, not a guarantee.

Where you distribute doesn’t change how you master. If you haven’t picked a distributor yet, the best music distribution services roundup compares seven of them, and the release hub covers the rest of release prep, from ISRCs to cover art.