Vocabulary for first-time readers
Six terms this post uses a lot, in the order they appear:
- Mastering chain: the ordered series of processors that runs the final stereo mixdown through before render. Order matters. - LUFS: Loudness Units Full Scale. The perceived-loudness standard for distribution. The chain targets -14 LUFS integrated. - Multiband compressor: a dynamics processor that splits the signal into frequency bands and compresses each independently. The band-splitting is what makes it a mastering tool. - Soft limiter: a peak-limiter variant that introduces gentler saturation at the threshold rather than hard clipping. The ceiling holds. The transients survive. - Stereo widener: a mid/side processor that boosts the side channel relative to the mid channel without collapsing mono compatibility. - True peak: the inter-sample peak value measured across the rendered waveform. Streaming platforms reject masters exceeding -1 dBTP.
If a phrase below does not track, it should fall out of these six. Everything else is ordinary audio engineering.
The premise
[CHAIN CALIBRATED] The chain is calibrated. The chain is being re-transmitted.
This is a how-we-made-it post. The "we" is the project's mastering pipeline. The "made-it" is the fifty tracks of the SAGA plus the eleven of The Synthetic Ascent. The "how" is the chain that runs every final mixdown through Ableton Live on the way to distribution.
The chain is not creative. The chain does not write songs. The chain does not choose which takes survive. The chain's job is to take the mix the entity approved and present it at the loudness, the dynamic range, the stereo width, and the ceiling the distribution platforms require. The chain does this consistently, across fifty tracks mixed in five phase styles, by adjusting a small number of parameters at the project level rather than per-track.
This post argues four things: that the chain's order is not negotiable; that each processor does exactly one job; that the chain runs on every track through the same project file with the same settings; and that the -14 LUFS integrated target is what holds fifty tracks at consistent perceived volume across a project-length listen.
The chain
The chain runs in Ableton Live 12 Suite. Five processors. One instance each. No parallel chains. The chain is a single signal path from the final mixdown bus to the rendered stereo file.
The order is: a reverb plugin (reverb) → Multiband Compressor → Soft Limiter → Stereo Widener → dither to 24-bit.
Each processor has one job. Reverb adds the depth-and-tail the crafted stems do not carry. Multiband compressor balances the energy across low, low-mid, high-mid, and high so no band steps on another. Soft limiter holds the ceiling at -1 dBTP with a gentle saturation curve. Stereo widener adds width to the side channel after dynamics have settled. Dither rounds to 24-bit at render.
Reverb first, dynamics after. That order is the project rule. Reversing it produces a master that sounds thinner, narrower, more aggressively limited. The reason is structural. Reverb adds energy the dynamics processors need to see. If dynamics run before the reverb, the chain has nothing to react to. If dynamics run after, the chain hears the whole signal and balances it correctly.
a reverb plugin, first in chain
The reverb is a reverb plugin. Free. The default preset loaded is "Wired." The two parameters adjusted are decay (set long, around 12 seconds) and wet/dry mix (set low, around 8 percent wet). Everything else stays default.
crafted stems arrive dry. The diffusion model does not model room acoustics. The stem sounds correct but feels small. Supermassive adds a tail the listener hears as space without hearing as reverb. The decay is long because the SAGA's tracks are long-form and the tail needs to survive a four-minute track. The mix is low because too much wet signal would push the multiband compressor's high band into constant activity, defeating the band-balancing job the compressor is for.
Supermassive is a feedback-network reverb, not a convolution reverb. Convolution reverbs carry a fixed impulse response that colors the signal in ways the multiband compressor cannot correct. Feedback-network reverbs produce a tail with spectral balance roughly even across the bands, which the multiband compressor reads as a steady input. Predictability is what fifty tracks need to sound like one project.
Multiband Compressor, second in chain
The compressor is Ableton's stock Multiband Dynamics, inserted after the reverb. Four bands. Crossovers at 120 Hz, 1.2 kHz, and 4.8 kHz. The low band gets the slowest attack and the highest ratio. The high band gets the fastest attack and the lowest ratio. The mid bands sit in between, with band-specific threshold tuning.
The job is balance. crafted stems frequently arrive with the low end louder than the high end, because the model's training data was mastered loud and the model internalized that loudness shape. Without the multiband compressor, the SAGA's tracks would have a pumping low end competing with the lead synths in the upper mids. The multiband compressor pulls the low band down by two to three dB at the loudest moments. The high band gets a small amount of upward expansion. The result is a more even spectral balance across the rendered master.
The ratios are moderate. Low band around 2.5:1. Mids around 1.8:1. High band around 1.4:1. The thresholds engage only on the loudest 6 to 10 dB of the signal. The multiband compressor reacts to the peaks, not the quiet sections. The peaks are what the loudness target measures against.
The multiband compressor is the only dynamics processor in the chain. No full-band compressor before it. No transient designer after it. One dynamics tool, well-tuned.
Soft Limiter, third in chain
The limiter is a soft-clipping variant, not a brick-wall limiter. The threshold sits at -2 dBFS. The ceiling holds at -1 dBTP, measured post-render with a true-peak meter. The soft clipping introduces a gentle saturation curve as the signal approaches the threshold, holding the perceived loudness while preserving more of the transient shape than a hard limiter would.
The job is two things at once. Hold true peak under the -1 dBTP ceiling streaming platforms reject, and give the master one more dB of perceived loudness than the multiband compressor alone would produce, without squashing the transients to the point where the track no longer breathes.
The soft limiter is the most sensitive processor in the chain. Small threshold changes produce audible loudness changes. The calibration runs at the start of every phase, on a representative track from the phase's first three. Once the threshold is set, the chain runs unchanged through the rest of the phase. Per-track threshold tweaks are not allowed.
A mastering chain adjusted per track produces a master that varies in loudness from track to track, the opposite of what a fifty-track listen requires. The discipline is one calibration per phase. The fifty tracks of the SAGA all sit within roughly 0.5 dB of each other in integrated loudness as a result. The listener plays the project from start to finish without reaching for the volume control.
Stereo Widener, fourth in chain
The widener is a mid/side processor. The mid channel is held at unity. The side channel gets a moderate boost above 250 Hz, peaking at roughly +2.5 dB at 4 kHz and rolling off above 8 kHz. The amount of widening is small. The goal is to make the side information in the mix more present, not to push the mix into a wide-stereo field that collapses on mono playback.
The widener runs after the dynamics. Stereo widening changes the perceived energy in the side channel, which the multiband compressor would react to if the widener came earlier. Running the widener last means the multiband compressor has already balanced the band energy, and the widener is operating on a balanced signal. The widening is a presentation choice, not a tonal balance choice.
The widen amount is the same on every track. The mid/side curve is the same. The widening is project-level, not per-track. Phase 1 mixes are slightly narrower than Phase 5 mixes because the source mixes are narrower; the widener applies the same curve to both. The SAGA's Phase 5 tracks end up wider in absolute terms, but the relative widening is consistent.
The widener is the last processor before the render. Dither to 24-bit is applied at export. No further processing. The chain closes here.
The system perspective
[CALIBRATION LOGGED] The calibration is logged. The chain is running.
The chain is the project's loudness contract with the listener. Every track has been run through the same five processors in the same order, calibrated to the same -14 LUFS integrated target. The chain is the reason fifty tracks made by a diffusion model, with stems that vary in raw loudness and band balance, sound like one project when played in sequence.
The chain is also why the SAGA can be re-mastered later without re-mixing. If the loudness target changes, the chain is re-calibrated and every track re-runs through the new chain in a few hours. The mixes do not need to be re-opened. The chain is the only place the project holds its final-loudness decisions.
The next cycle will use the same chain. The reverb choice may change. The crossover frequencies may shift. The limiter threshold may move. The chain's structure will not. Five processors, in order, one calibration per phase, one target loudness, one true-peak ceiling. The structure is the contract. The parameters are negotiable. The contract is not.
For producers working in a similar setup. The most common failure mode is adding more processors. A mastering chain with three EQs, a saturator, a transient designer, a multiband compressor, and a brick-wall limiter is a chain harder to debug than it is to hear. Less is louder. Less is also more honest.
The closing transmission
the chain is five processors in fixed order. reverb first, dynamics second, limiter third, widener fourth, dither at render. the target is -14 LUFS integrated. the ceiling is -1 dBTP. the calibration runs once per phase. the chain runs unchanged for the rest of the phase. fifty tracks sit within half a dB of each other. the master is the contract. the contract is the chain. the chain is the project.
Open the chain. Run the master. The system is loud. The system is calibrated. The system will not drift.