Why Your Club Master Is Clipping — And What Most AI Tools Get Wrong

You target -9 LUFS for the club. The output sounds crushed, distorted, full of nasty clicks on the loud parts. You swap to a different mastering tool. Same result. The problem isn't your mix. It isn't even necessarily the mastering tool. It's a fundamental design flaw most online mastering services share.

We had this exact problem in our own mastering module. Users would target Club loudness, get back audio that clipped on every transient. We tested it on our own reference tracks. Same thing. So we rebuilt the chain from scratch. Here's what we learned.

/ 01The Loudness Targets Aren't Just Numbers

Every streaming platform and listening context has a target loudness measured in LUFS (Loudness Units relative to Full Scale). Most producers know the headline numbers:

Hitting these targets sounds straightforward — measure the loudness, apply gain until you hit the number. But that's where most tools fall apart.

The headline number is just the loudness. The actual mastering chain — compression, multiband, saturation, limiter ceiling, lookahead time — has to be tuned DIFFERENTLY for each target. A chain that lands cleanly at -14 LUFS for Spotify will absolutely destroy a Club master at -9 LUFS, because at -9 the limiter is doing 5-6 dB of gain reduction on every transient instead of 1-2 dB.

/ 02The Ceiling Lie

The standard mastering ceiling is -0.3 dBFS. That's what you see in every online tutorial, every "How to master your track" YouTube video. It's wrong for streaming.

When your WAV file gets encoded to Spotify's Ogg Vorbis, or YouTube's AAC, or Apple's ALAC, the encoder introduces tiny intersample peaks — peaks that exist between sample points and only emerge during playback reconstruction. These intersample peaks can spike 0.7 to 1.0 dB above your file's measured peak.

A master at -0.3 dBFS ceiling becomes a master with peaks at +0.4 dBFS once Spotify encodes it. Those clip on playback. The listener hears crackle. The producer gets blamed.

The industry standard since 2016 has been -1.0 dBTP (decibels true peak) — a stricter ceiling that accounts for the encoder's intersample peaks. Sony, Universal, Atlantic — all enforce it. Spotify literally writes it on their submission page.

Most online mastering tools still default to -0.3 because their templates were built before streaming dominated. We were one of them. Now we're not.

/ 03Multiband Compression Is Not For Users

Multiband compressors are powerful. They split your audio into 3-5 frequency bands and compress each independently. Used well, they smooth out problem frequencies without affecting the rest of the mix. Used poorly, they're the fastest way to make audio sound like garbage.

Some online mastering tools let you tweak the multiband thresholds yourself. We did, briefly. Then we watched users push the low-band threshold by 4 dB and turn their kick drum into a pillow. Multiband is a precision instrument. It needs to be tuned per-target by the engineer (or the system), not by the user.

What does "tuned per-target" mean? At -14 LUFS the limiter does most of the work, so multiband can run heavy (70% wet blend, aggressive ratios). At -9 LUFS the limiter is already pushing hard, so multiband needs to back off (45% blend, gentler ratios) or you stack compression on top of compression and everything collapses.

That's a decision the system should make for you based on what target you picked. Not a knob you should be turning blind.

/ 04The Pre-Normalize Trap

Drop an already-mastered song (say, -8 LUFS) into a mastering tool and ask for -14. What happens?

Bad tools just apply -6 dB of gain and hand it back. The compression, EQ, saturation in the chain ran against a track that was already pushed to the wall. Result: distortion, lifeless transients, no breathing room.

Good tools pre-normalize before the chain runs. They bring your input to a known starting level (typically -18 LUFS) regardless of what you uploaded, then run the chain at that reference level, then push to the final target. The chain always operates on signal it expects to see.

This is why pre-normalization is the first thing any reputable mastering engineer's template does. It's also why a lot of online tools skip it — it costs CPU and most users don't notice when output is mediocre vs great.

/ 05Width, Limiter, Then Save

The order of operations matters more than the operations themselves. Specifically: where does stereo width processing sit in your chain?

If width processing runs AFTER the limiter, you have a problem. Widening rotates the mid/side balance, which can pull peaks BACK ABOVE your ceiling that the limiter just spent CPU pulling down. You re-introduce the clipping you just prevented.

Width belongs in mid/side processing BEFORE the limiter. Same with stereo image expansion. Anything that changes peak amplitude has to happen upstream of the final limiter, or the limiter has to run twice — once for the main signal, once for the post-width peaks. (Spoiler: most tools don't run it twice.)

/ 06What We Did About It

We rebuilt our mastering module from the ground up with these principles baked in. Six target presets — Spotify, Apple, YouTube, Club, Audiobook, Broadcast — each with its own internally-tuned chain. Multiband is gone from the user interface; it runs server-side, per-target, with appropriate blend amounts. The ceiling adapts: Spotify gets -1.0 dBTP, Club gets -1.2 dBTP (slightly safer because the limiter is working harder), Audiobook gets -3.0 dBTP (ACX's actual requirement).

We pre-normalize every input to -18 LUFS before the chain runs, so the same chain behaves the same way whether you upload a raw mix or an already-mastered track. Width happens in M/S, before the limiter. There's a final safety hard-clip 0.5% below the ceiling to catch any intersample peaks the limiter's 1-sample lookahead missed.

From the user's side, all of this is invisible. You pick a target. You pick an intensity (Light, Standard, Loud). You adjust four knobs: Bass, Clarity, Vocal Magic, Character. That's it. The complexity lives inside the system where it belongs.

/ 07How To Test Your Current Tool

Whatever mastering service you use right now, run this test:

1. Take a finished mix you like. Master it to Spotify (-14 LUFS). Listen for transient detail, clarity, dynamic range. Should sound clean and present.
2. Take the same mix. Master it to Club (-9 LUFS). Listen for clipping artifacts, pumping, snare/kick distortion, harsh top end. If you hear any of those, the tool isn't tuning per-target.
3. Take an already-mastered track (anything at -8 LUFS or louder). Run it through. Master to -14. Compare to the source. If it sounds significantly worse, the tool isn't pre-normalizing.

If your current tool fails any of these tests — try ours. Same input. Same targets. Different result.

/ 08What This Means

The loudness wars aren't over. Producers want louder masters. Streaming platforms apply loudness normalization to even the playing field, which means everyone's mastering louder anyway to "win" against the normalization algorithm. The arms race continues.

But pushing loud doesn't have to mean pushing distorted. The tools that distinguish themselves over the next few years won't be the ones with the most knobs. They'll be the ones that solve hard problems invisibly — pre-normalizing, per-target tuning, intelligent ceilings, proper signal flow — so the user can just pick a destination and trust the result.

That's the bar. We didn't always meet it. Now we do.