Why multi-track releases drift apart
A mix decides how a track sounds; mastering decides how loud and polished it is next to other music. When each track is finished at a different time, with different settings, or by a different tool — or simply never mastered at all — each one lands at its own loudness, with its own tonal character. The differences are invisible while you work on songs one at a time. They only become audible when the songs sit side by side.
AI-generated music makes this drift worse, not better. Every Suno or Udio generation is an independent event: a dense, wall-of-sound rock generation comes out with a completely different energy level than a sparse piano ballad, and even two generations of the same prompt can land at noticeably different levels and brightness. String six independent generations into an EP and you've effectively hired six different engineers who never spoke to each other.
"Won't Spotify normalize it anyway?" — no, not the way you hope
This is the most common objection, and it's worth being precise about. Spotify does apply loudness normalization by default — but for album playback, its official documentation states: "We normalize an entire album at the same time, so gain compensation doesn't change between tracks" (Spotify's loudness guidance). In plain terms: when someone plays your EP top to bottom, Spotify turns the whole release up or down as one unit. If track 5 is 4 dB quieter than the rest, it stays 4 dB quieter. Per-track adjustment only kicks in when your songs are shuffled or dropped into playlists.
That's the right behavior, by the way — it's what preserves intentional dynamics, like a quiet interlude, on professionally mastered albums. But it means the platform trusts you to have set the relative levels on purpose. And that's before considering listeners who have normalization switched off, downloads, DJ software, and platforms with different rules. Consistency across a release is a mastering job, and it's yours.
The fix: master the release, not the tracks
The principle that solves this is old and simple: an EP or album should be mastered as a set — same tool, same approach, one sitting. You don't need a mastering suite to apply it. Here's a workflow that works entirely without a DAW.
Step 1: Finish everything first. Don't master track 1 the week you make it and track 6 a month later. Wait until every track is final, then master them together in one session. Half the consistency problems on self-released EPs come from mastering across weeks with drifting settings and drifting ears.
Step 2: Start from the best source files. Export or download the highest-quality version of each track — WAV rather than MP3 where you have the choice. Mastering amplifies what's in the file, including compression artifacts you can't remove later.
Step 3: Run every track through the same mastering pass. Same tool, same settings (with an automatic tool: just the same tool — that's the point). Processing all six back to back, in one sitting, is what gives the release a shared character instead of six private ones.
Step 4: Listen to the seams. This is the step almost everyone skips. Make a playlist of the mastered tracks in release order and audition the transitions: the last 30 seconds of each track into the first 30 seconds of the next. You're asking one question — does the next track feel like it belongs to the same record? A small deliberate contrast (a ballad sitting lower) is fine. A jump that makes you reach for the volume knob is not.
Step 5: Fix outliers at the right layer. If one track sticks out, re-master that track first — a different pass may suit it better. If it still sticks out, the problem is usually in the source: a generation that came out harsh or thin will fight any master, and re-generating is faster than repairing. (Our earlier posts cover how to judge a single master by ear and why AI tracks need mastering at all.)
Step 6: One full pass on small speakers. Play the whole release start to finish on phone or laptop speakers. Ten to twenty minutes. If it survives as one record there, ship it.
What "consistent" actually means
Consistency is not "every track hits exactly −14 LUFS." Chasing an identical number across a ballad and a banger produces a flat, lifeless release — and Spotify's own guidance treats −14 LUFS as a reference point for masters, not a law (source). What you're after is perceived consistency: no track makes the listener adjust the volume, and no track sounds like it wandered in from a different session. Quiet songs may sit lower on purpose. The test is your ears on the transitions, not a meter chasing a number.
Where Tonekai fits
We build Tonekai, a local AI mastering app for macOS, and multi-track releases are where its design choices pay off. Every track gets the same automatic analysis and processing — no settings to accidentally change between track 2 and track 5 — which is exactly the "same tool, same approach, one sitting" this workflow calls for. Because everything runs on-device (no uploads, Apple Silicon native, macOS 14+), mastering six tracks and re-doing two of them is a minutes-long loop, not an afternoon of upload queues. And the built-in A/B lets you check each track against its original before you export. The app is free to try, with a one-time Pro Unlock — no subscription — for WAV/AAC export and unlimited mastering.
Download Tonekai on the Mac App Store — free to try.
Frequently asked questions
Should every track measure the same LUFS? No. Matching perceived level at the transitions matters more than matching a number. If two adjacent tracks feel wrong, adjust; if a quiet track feels intentional, leave it.
Do I need to master the whole EP as one long file? No — that's a vinyl-era technique that mostly adds hassle for digital releases. Master per track, but in one session with one tool, then verify the sequence by listening.
Does track order matter for this? Yes, more than people expect. A big jump between tracks 3 and 4 sometimes disappears if a mid-energy track sits between them. If one transition refuses to behave, try re-sequencing before re-mastering.
I'm releasing singles, not an EP. Does any of this apply? The "same tool, same approach" principle still pays off: your singles form a de-facto catalog, and listeners hit them back to back on your artist page and playlists — where per-track normalization helps, but tone consistency is still on you.
The bottom line
A release that hangs together is mostly a scheduling decision: finish everything, master everything the same way in one sitting, then listen to the seams and fix the outliers. The platforms won't equalize your album for you — album normalization deliberately preserves the differences you shipped. Ship differences you chose.