Preparing your premaster for mastering

Whether the master is being cut by an engineer, an algorithm, or future-you next weekend, the quality ceiling is set by the bounce you hand over. A clean premaster takes ten minutes to prepare. A compromised one quietly costs you punch, depth, and low-end control that no mastering stage can give back.

What a premaster is

A premaster is the final stereo bounce of your finished mix, exported without the mastering processing — the file mastering starts from. Everything that is part of the music stays in: your sound design, bus compression that's part of the groove, saturation on the drum bus. What comes out is anything whose only job was to make the mix loud or "finished" on your monitors.

Take the limiter off the master bus

This is the one that matters most. A brickwall limiter on your 2-bus flattens the very transients and dynamic headroom that mastering needs to work with. Once a peak has been squashed into the ceiling, the information is gone — a mastering limiter downstream can only crush it further, never restore it. Deliver the premaster with the limiter (and any "maximizer" or clipper doing loudness duty) bypassed.

Two honest nuances:

• If a limiter or clipper is genuinely part of your sound — a clipped drum bus that defines the track's aggression — keep that, because removing it changes the music. The rule targets the final safety/loudness limiter on the master, not creative processing inside the mix.
• Mixing into a limiter is fine; exporting through it isn't. Many producers write and mix with a limiter on to feel the loud version. Great — just bypass it for the bounce, and glance at the mix balance afterward to confirm nothing was leaning on it.

Headroom: how much and why

The convention is to leave roughly 3–6 dB of peak headroom, with peaks landing somewhere around −6 to −3 dBFS. The number itself is not magic — in a 24-bit or 32-bit float export there is no fidelity lost at −6 versus −1. The headroom exists for practical reasons:

• It proves nothing upstream is clipping — a mix that "peaks at 0.0" usually means intersecting overs you can't see.
• It gives mastering EQ room to boost a band without immediately slamming the ceiling.
• Some processors (analog emulations especially) are calibrated to expect sensible operating levels.

Do not normalize the export, and do not turn a hot mix down with a gain plugin after clipping has already happened. If your mix bus is clipping, lower the channel or bus levels feeding it — pulling the master fader down post-clip just gives you a quieter copy of the damage.

Sample rate, bit depth, format

• Sample rate: export at your project's native rate. If you produced at 44.1 kHz, deliver 44.1; at 48, deliver 48. Upsampling a 44.1 session to 96 kHz adds no information — it just makes a bigger file. Downsampling before mastering throws away the chance to do that conversion once, properly, at the end.
• Bit depth: 24-bit, or 32-bit float. Never 16-bit for a premaster — 16-bit is a final delivery format. 24-bit gives you ~144 dB of theoretical dynamic range; your headroom costs nothing there.
• Format: WAV or AIFF. Never MP3 or AAC — lossy encoding permanently discards audio, adds its own inter-sample peak problems, and cannot be undone by mastering.
• Dither: not needed. Dither applies only when reducing bit depth to a final format (like 16-bit). A 24-bit premaster headed into mastering should be exported clean; the mastering stage handles dither on the final render.

Heads, tails, and edits

Export from the very start of the track to at least a second past the last reverb tail — let effects ring out fully rather than cutting the bounce at the final bar line. Don't fade the intro or outro "to be safe"; fades are a mastering decision and easier to add than remove. And listen to the whole render top to tail once before sending it anywhere: render glitches, a soloed channel, a muted return, or an automation jump caught at export time costs a minute; caught after mastering, it costs the whole pass.

The checklist

• Master-bus limiter / maximizer / loudness clipper bypassed
• Peaks around −6 to −3 dBFS, no clipping anywhere in the chain
• No normalization on export
• WAV or AIFF, 24-bit (or 32-bit float), project's native sample rate
• No dither applied
• Full length exported, reverb/delay tails ring out, no safety fades
• Rendered file auditioned start to finish
• Mono check passed — the mix doesn't collapse (critical for the low end)

Hand over a bounce like that and every downstream decision — the reference match, the loudness push, the true-peak ceiling — starts from full information instead of fighting damage.