Mastering to -14 LUFS: why every SONICHAOS track ships pre-targeted

Every track in the SONICHAOS catalog ships at -14 LUFS integrated, with the true-peak ceiling sitting at -1.0 dBTP. The number is not arbitrary, and the chain that gets us there is not a secret. This post walks the full path from raw stems to the WAV you download, explains why the target is what it is, and ends with a small list of edits you should not redo on import.

What "-14 LUFS" actually means

LUFS stands for Loudness Units relative to Full Scale. The integrated measurement averages the loudness of the entire programme, gated against silence so a quiet intro does not pull the number down. Spotify, Apple Music, YouTube, and TikTok all normalise to a target near -14 LUFS for most listeners. Push your master to -8 and the platform turns it down. Land at -22 and the platform pushes it up, surfacing tape hiss and pre-amp noise that you would rather keep buried.

The practical consequence: there is no loudness war on streaming. There is only a target, and a true-peak ceiling that protects against inter-sample peaks on the consumer's lossy decoder.

The chain we run

Each contributor's bounce arrives as a 24-bit, 48 kHz WAV. The lab pipeline runs three passes on that file before it lands in the catalog:

1. DeepFilterNet 3 for pre-clean. The model removes hum, hiss, and the low-grade rumble that creeps in from quiet rooms. Default sensitivity is 12 dB of attenuation; we cap it at 6 dB to avoid pulling the noise floor into the music.
2. Matchering 2.0 for reference matching. The reference is one of nine curated targets keyed to the track's top genre tag. Matchering nudges the spectral envelope and the dynamics toward the reference without re-EQing the lead element to death.
3. An internal limiter at -1.0 dBTP ceiling, set to a 20 ms release and a 1 ms lookahead. The limiter is not the loudness sculptor. It only catches the peaks Matchering missed.

Integrated loudness lands at -14.0 LUFS ±0.3. Short-term loudness is allowed to swing freely; we do not flatten dynamics for their own sake. A trailer cue still hits.

What we never do

Three rules govern the chain:

• No multi-band compression on the master bus. If a track needs that, the fix is at the contributor's mix stage.
• No mid/side EQ above 8 kHz. The high-frequency stereo image is the composer's signature; we do not paint over it.
• No loudness above -13.5 LUFS, even when the reference is louder. Streaming will turn it down anyway.

The combination matters. You can run any one of these three rules and still ship a busy master. Running all three gives you headroom for transients without sacrificing the perceived loudness budget.

The signed PDF receipt

Every download writes a license receipt as a PDF, ed25519-signed against the asset_id and the time of download. The receipt records the loudness target and the ceiling. If a client asks why a cue sounds quieter than a reference they sent you, the receipt is the audit trail.

What to leave alone on import

This is the part most working creators get wrong: they treat a SONICHAOS track like a raw stem and slap a master-bus chain over it. Don't.

• Skip the limiter. Adding another peak limiter on top of ours pushes the inter-sample peaks back above 0 dBTP on consumer decoders.
• Skip the loudness maximiser. The track is already at platform target. Anything you add is fighting the platform's normaliser.
• Skip broadband saturation. It changes the spectral balance Matchering matched against. Reach for narrow-band EQ instead, in the 2.5–4 kHz range, if the cue is sitting too far behind your dialogue.
• Re-EQ for context, not for taste. A film cue under a voiceover often needs a -3 to -4 dB notch around 2.5 kHz to clear the speech band. That is a context edit, not a master change.

If you absolutely have to lift the cue under a louder bed, group the cue with the rest of your music and master the group, not the cue alone.

When the target changes

Two cases pull the target somewhere other than -14 LUFS:

1. Theatrical delivery. Cinemas spec -27 LUFS integrated, with peak headroom up to -2 dBTP. The SONICHAOS cinema bundle re-renders contributor masters at this target on demand.
2. Broadcast TV. EBU R128 specifies -23 LUFS integrated, +/- 0.5. The broadcast bundle uses the same chain with a different limiter ceiling and integrates the target through the entire programme.

Both bundles are flagged on the track detail page. The default download is streaming-targeted; the alternates are one click away.

A short FAQ

Will Spotify still turn my video down? Spotify normalises audio in the Spotify app. YouTube normalises in the YouTube player. Your video on someone else's site is at the loudness you mastered it to. Match the bed to the dialogue level your editor uses, then trust the platform.

Do I need a true-peak meter? Yes, if you are mastering anything yourself. The free options are good enough: youlean Loudness Meter, Voxengo SPAN Plus. Set the ceiling at -1.0 dBTP and the integrated target at -14 LUFS for streaming.

What about loudness for ads? US TV ads spec CALM Act compliance: -24 LUFS integrated, +/- 2. Online video ads on most platforms target somewhere between -16 and -14 LUFS. Spec varies by platform; the safest approach is to match the surrounding content's level rather than chase a number.

The catalog page for each track lists the loudness, the ceiling, the sample rate, and the bit depth. None of those need to be guessed.

Browse the catalog and the loudness target is on every detail page.