rekordbox Auto Gain vs MP3Gain: Trust the Software or Fix the Files?

Every major DJ application now levels your tracks for you — and every one of them does it differently, keeps the result in a different place, and forgets it the moment you step outside that app. MP3Gain takes the opposite bet: change the file once, losslessly, and the leveling travels with the music forever. Neither approach is magic. This page explains what each auto-gain system actually does, so you can decide which layer deserves your trust.

What rekordbox Auto Gain actually does

When rekordbox analyzes a track, it calculates an auto-gain value (shown as A.GAIN) intended to bring the track to a consistent playback level, and stores that value with the track in the rekordbox collection. With the Auto Gain preference switched on, the value is applied as you play, and you can fine-tune it per track in the grid editor. Two limits matter. First, the value lives in rekordbox’s world, not in the audio — play the same file in another application or a media player and it is exactly as uneven as the day you downloaded it. Second, Pioneer’s own support forum states plainly that auto-gain data is not stored in exports: prepare a USB for standalone club gear and the leveling work you did inside rekordbox does not ride along. The file is the only thing the CDJ ever sees.

The other three, briefly

Serato DJ. Serato’s manual describes its Use Auto Gain option this way: during the overview-building (analysis) process it calculates an auto-gain value for each track — the adjustment needed to match the track’s perceived loudness to a reference level you choose, from 89 dB (quieter) to 98 dB (louder), with 92 dB the recommended default. The value is applied each time the track loads, by setting the track gain knob. Note what that means: Serato’s reference is its own, its default is 92 dB rather than anyone else’s number, and the result is a playback setting inside Serato’s ecosystem.

Traktor. Native Instruments’ manuals describe Autogain as relying on gain values extracted when a track is analyzed on import; a “Set Autogain When Loading Track” preference applies that stored value as the track loads, and the auto-gain value is kept separately from the channel gain you set by hand. Same pattern: analysis inside the app, value inside the app’s data, silence everywhere else.

VirtualDJ. VirtualDJ’s Auto Gain normalizes a track toward a 0 dB level for playback, and its default “Auto+Remember” mode additionally remembers gain tweaks you make by hand. A 0 dB target is a peak-style goal — a different philosophy from the perceptual loudness matching above, and one more reason the four apps do not agree with each other about how loud a track “should” be.

The disagreement, tabulated

“But doesn’t the software write something into the file?”

Sometimes, yes — and it changes nothing about the argument. Serato’s manual, for example, notes that a track gain adjustment made in the software is saved to the file and retained for the next load. What is saved is Serato’s own gain value, in Serato’s own terms: a playback instruction that Serato reads. rekordbox does not apply it, your media player does not apply it, a CDJ reading the bare file does not apply it — and crucially, it is not a change to the audio itself, so the file’s actual loudness, the thing every other playback chain responds to, is exactly what it was. A note in one app’s handwriting, stored in the file’s pocket, is not the same as the file being level. MP3Gain’s adjustment is the opposite arrangement: no app-specific instruction at all, just an audio file that is measurably at the shared level for anything that plays it.

The same distinction explains a common confusion: a track can carry a pocketful of metadata — one app’s gain tag, another’s analysis markers — and still blast or whisper the moment it is played somewhere that reads only the audio. When DJs say leveling “didn’t stick,” it is almost always because the leveling they did lived in that metadata layer. File-level gain is the layer that cannot fall off.

When the software’s auto gain is genuinely enough

If you play on one laptop, in one application, and never export to standalone gear, auto gain is convenient and mostly effective: the app re-levels on load and you touch nothing. It is also the only option for formats MP3Gain cannot touch. For a single-app, single-machine DJ, this page is not an argument to change your life — it is an explanation of the boundary you are inside.

When file-level leveling wins outright

The moment any of these is true, the file is the only common denominator: you export USBs for CDJs or XDJs; you keep a backup laptop with different software; you share tracks with a B2B partner; you play the same files in a media player, a car, or a warm-up stream; or you simply want the leveling decision to outlive whichever app you happen to use this year. Software comes and goes — DJs have already buried several industry-standard applications. The MP3 file, leveled once, does not care.

There is also a quieter asymmetry worth noticing: auto gain is computed when a track is analyzed, from the file as it exists at that moment. Add a track in a hurry before a gig and, depending on the application and your settings, it may be played with provisional or missing analysis — the one situation where you most need the help is the one where the database has not caught up. A leveled file needs no catching up: its level is a property of the audio itself, true on first load, on any machine, including machines that have never heard of your library. Reliability that depends on a database staying in sync with a folder of files is a different, weaker kind of reliability than a file that is simply right.

The sensible combination

These are layers, not enemies. Level the files with MP3Gain first (the full workflow is MP3Gain for DJs), then let your DJ software re-analyze: its auto gain now starts from files that are already close, so it only ever applies small corrections instead of heroic ones. Two cautions. Do not chase loudness twice — if a track still feels quiet after both layers, the honest answer is that it was mastered quiet, and the trim knob exists for the last decibel or two. And re-analyze after leveling rather than trusting stale analysis values; your cue points are safe through all of this, for the reasons in the cue point explainer. If your real requirement is loudness measured to the modern LUFS standard across several formats, that is a different tool class entirely — see MP3Gain vs waveAlign and LUFS for DJs.