Do I automate before or after compression?

After, nearly always. Automating before changes what the compressor receives and therefore how much it compresses: you push a verse up and the compressor pulls it back down, so you end up fighting your own automation. With the compressor already set, automation acts on the result and does exactly what you ask. The exception is corrective automation — evening out individual syllables in a vocal — which does go in front, so the compressor has less work to do.

How much level counts as an audible change?

Less than people assume. In a dense mix, 0.5 dB on a specific element already reads as «that came forward», and 1 dB is an obvious change. The 3 dB jumps beginners draw usually sound like a mixing error. If you need 6 dB for something to be heard, the problem isn't level: something else is masking it and that gets solved with EQ or with the arrangement.

Should I automate the fader or a gain plugin?

The channel fader for the song's balance, and a gain plugin at the top of the chain for corrective automation. Keeping them separate has a big practical payoff: you can redo the whole balance without destroying the fine syllable-levelling work, and vice versa. Putting both on the same fader turns any change of mind into a lost afternoon.

What do I automate first if I'm short on time?

The lead vocal, by a distance. It's the element the listener follows and the only one where a lost syllable registers as a mistake. After that, the reverb and delay sends, which add air in the empty sections and clutter the full ones. The rest of the mix survives static faders reasonably well; the vocal doesn't.

Is it wrong to compress instead of automating?

Not wrong, but it does something different. A compressor reacts to every peak equally, with no idea whether that syllable matters; you do know. Compressing heavily to tame a huge dynamic range also flattens what you wanted to keep. The usual arrangement, and the one that works best, is to automate first to get the signal into a sensible range and let the compressor handle only the fine adjustment that's left.

Is automation worth it on a demo?

The vocal, at least, yes. It's what makes a demo sound like a song rather than a rehearsal, and it costs very little: four or five moves per track. What isn't worth it on a demo is fine send and pan automation, because it tends to change completely the moment the arrangement does.

Mixing10 min read27 August 2026

By Mónica Andrade

Volume automation: the step between a static mix and a finished one

Why no static fader works for a whole song, what to automate before anything else, and how to do it without spending the afternoon drawing ramps.

Why no static fader works

A song isn't a state, it's a journey. The space a vocal needs in a bare verse isn't the space it needs when the chorus guitars come in, and the fader setting that solves one ruins the other. When someone says their mix «sounds good in places», they're almost always describing exactly this: they found a setting that works in one section and left it there for three minutes.

That's why automation isn't a finishing flourish or an advanced stage: it's what turns a balance into a mix. And it's why it pays off so fast — it's the only process in the whole chain that can make different decisions at different moments, which is exactly what a listener does.

The two automations, which aren't the same thing

Nearly all the confusion around this comes from lumping together two jobs that have nothing in common. It's worth separating them from the start, because they happen at different times, with different tools and by different criteria.

	Corrective	Mix
What it fixes	Uneven syllables and notes inside a take	How much weight each element has in each section
Where it goes	Gain at the top of the chain, before the compressor	Channel fader, after everything
Scale	Tenths of a second, 1–3 dB	Whole bars, 0.5–2 dB
When you do it	Before mixing anything	Once the base balance works
You notice it missing when	The vocal «drops out» on single words	Every section sounds equally full

Corrective automation gets done once and never touched again. Mix automation changes every time you change your mind about the song. Keep them in separate places and changing your mind costs ten minutes; put them on the same fader and it costs you the lot.

Where to start

The order matters because each step changes the criteria for the next. This one works in any style and doesn't depend on which DAW you use:

Even out the lead take before mixing: bring up the words that get lost, pull down the ones that jump. That alone lets whatever compressor you put after it work far more gently.
Mix the densest section of the song — usually the last chorus — with the faders static. It's where there's least room, so if something fits there it fits everywhere.
Listen to the whole song and note where something gets in the way or goes missing, without touching anything. Listening and deciding at the same time is the fastest way to get stuck on one bar.
Work through those notes one at a time and solve each with the smallest move that works: nearly always between half a decibel and two.
Automate the reverb and delay sends last: more tail in the empty parts, less in the full ones. It's the adjustment that buys the most depth per move made.

That first step — evening out the take — goes faster if the vocal chain is already sorted; that's in the guide on mixing vocals. And if automating the sends makes the reverb muddy rather than airy, the problem is usually further back, in how the effect is set up.

The mistakes that cost the most time

Drawing ramps by hand when riding the fader while it plays would have done: the second nearly always sounds more musical, because it's the ear making the call rather than the eye.
Automating with the compressor unset: you push the verse up, the compressor puts it back, and you end up adding 6 dB to gain 1.
Big moves: if you need more than 3 dB, what's failing is the balance or the arrangement, not the automation.
Automating every channel as a matter of course. A normal mix needs moves on three or four tracks; the rest stay put and nothing bad happens.
Doing it while watching the waveform. Automation is decided by hearing the whole song, not by looking at peaks.
Not letting the session rest before the final pass: moves that felt subtle yesterday sound like steps today.

Riding a fader with judgement while the music plays is a manual skill, and it trains separately from the analytical ear. Gain Rider is precisely that exercise: following a signal that keeps moving without overshooting or lagging behind.

Frequently asked questions

Do I automate before or after compression?: After, nearly always. Automating before changes what the compressor receives and therefore how much it compresses: you push a verse up and the compressor pulls it back down, so you end up fighting your own automation. With the compressor already set, automation acts on the result and does exactly what you ask. The exception is corrective automation — evening out individual syllables in a vocal — which does go in front, so the compressor has less work to do.
How much level counts as an audible change?: Less than people assume. In a dense mix, 0.5 dB on a specific element already reads as «that came forward», and 1 dB is an obvious change. The 3 dB jumps beginners draw usually sound like a mixing error. If you need 6 dB for something to be heard, the problem isn't level: something else is masking it and that gets solved with EQ or with the arrangement.
Should I automate the fader or a gain plugin?: The channel fader for the song's balance, and a gain plugin at the top of the chain for corrective automation. Keeping them separate has a big practical payoff: you can redo the whole balance without destroying the fine syllable-levelling work, and vice versa. Putting both on the same fader turns any change of mind into a lost afternoon.
What do I automate first if I'm short on time?: The lead vocal, by a distance. It's the element the listener follows and the only one where a lost syllable registers as a mistake. After that, the reverb and delay sends, which add air in the empty sections and clutter the full ones. The rest of the mix survives static faders reasonably well; the vocal doesn't.
Is it wrong to compress instead of automating?: Not wrong, but it does something different. A compressor reacts to every peak equally, with no idea whether that syllable matters; you do know. Compressing heavily to tame a huge dynamic range also flattens what you wanted to keep. The usual arrangement, and the one that works best, is to automate first to get the signal into a sensible range and let the compressor handle only the fine adjustment that's left.
Is automation worth it on a demo?: The vocal, at least, yes. It's what makes a demo sound like a song rather than a rehearsal, and it costs very little: four or five moves per track. What isn't worth it on a demo is fine send and pan automation, because it tends to change completely the moment the arrangement does.