Does MP3Gain Mess Up Cue Points and Beatgrids? (rekordbox & Serato)

Short answer: no — and the reason is mechanical, not a matter of trust. Cue points and beatgrids are positions in time. MP3Gain never re-encodes your file, never touches its timing, and never changes its length, so there is nothing for a time position to drift away from. The fear is understandable, because other “normalizers” genuinely can shift your audio. This page explains the difference, what MP3Gain does change, and the one housekeeping step worth doing afterwards.

Why cue points are time, and time is untouched

When you set a hot cue at 1:32.400 or a beatgrid anchored to the first downbeat, your DJ software stores a position measured against the track’s timeline. Anything that preserves the timeline preserves every marker on it. MP3Gain’s entire design is timeline preservation: instead of decoding the MP3 to audio, changing the volume and encoding a new file, it adjusts the gain value stored inside each existing MP3 frame — the digital equivalent of writing a different number on the volume knob of every frame, in place. The audio data, the frame sequence, the duration and the timing are exactly what they were. A cue at 1:32.400 points at the same kick drum before and after, because the kick drum never moved.

Why the fear exists: re-encoding tools are a different story

Many volume tools — and every general audio editor used carelessly — work by decoding and re-encoding. Encoding an MP3 introduces encoder delay and padding at the start of the file, and a decode–edit–encode round trip can shift the audio by small amounts, rebuild the file’s structure, and rewrite its tags from scratch. Small shifts are exactly what break tight beatgrids and loops. MP3Gain was written in 2001 specifically to avoid this: no decode, no encode, no generation loss, no shift. If a tool cannot describe its method as “adjusting the gain field in each frame,” it is not doing what MP3Gain does.

What MP3Gain does change — honestly

Three things, none of them timing. First, loudness itself: that is the point of the tool. Second, tags: MP3Gain writes its analysis and undo information into the file’s tags (APEv2, or ID3v2 depending on settings), which is how its changes stay reversible years later. Tag-writing is a metadata operation; it does not move audio. Some older players have quirks reading APE tags — if you ever meet one, the symptom is tag display oddities, not broken cues, and our troubleshooting hub covers the tag-related fixes. Third, your DJ software’s stored picture of the track: analysis data such as auto-gain values and overviews were computed from the old loudness. They are not wrong about where things are; they are stale about how loud things are.

The one step to run afterwards: re-analyze

After a leveling pass, let rekordbox or Serato re-analyze the changed files. Re-analysis refreshes the loudness-derived data — auto-gain values and waveform overviews — so the software’s model matches the files’ new levels. Where your software keeps its analysis in its own database, that database is untouched by MP3Gain and simply gets its loudness entries updated on re-analysis; where data is kept with the track, the same refresh applies. Either way, the division of labour is clean: MP3Gain changes how loud the file is; re-analysis teaches your software the new fact. Cues and grids, being time data rather than loudness data, come through the whole process unchanged — but do the five-minute verification below anyway, because trust in a library is earned per workflow, not per article.

“My waveforms look different afterwards — did something break?”

Probably not, and the picture is worth understanding. The waveform overview your software draws is a loudness picture: it renders how much energy the track has over time. Level a quiet track up by 5 dB and its overview will be redrawn — after re-analysis — visibly taller, because the track genuinely is louder now. That redraw frightens people into checking their grids, which is healthy, but the thing to check is position, not height: the peaks and silences in the new picture sit at the same times as in the old one, because they are the same audio. Height is loudness; position is time; MP3Gain changed the first and cannot change the second. If an overview ever does appear shifted — a drop landing visibly earlier or later than your cue — that is the signature of a file that was re-encoded or replaced somewhere in its history, not of a gain adjustment, and the cure is to go back to the unleveled backup copy you made in step one of the protocol below and compare the two files directly.

One more reassurance for the export-minded: DJs who prepare USBs worry about the copy on the stick rather than the copy in the library. Leveling happens to the library files themselves, so any export made after the pass carries the leveled audio — the export contains the same files, with the same timing, at the new level. Export first and level afterwards, and you would simply export again; nothing about the workflow locks you in.

The safe first-run protocol

1. Copy one crate folder — twenty or thirty tracks you know intimately — and level the copy with the workflow in MP3Gain for DJs. 2. Load three of those tracks in your DJ software and jump through every hot cue: each should land exactly where your muscle memory expects. 3. Check one tight loop and one track where you set the grid manually. 4. Only then run the full library, using the backup-and-batch routine in how to batch normalize a large MP3 library. If anything ever looks off, MP3Gain’s undo restores the original gain from its tag data — the experiment is reversible by design.

And loops, memory cues and saved markers?

The same logic covers every marker type, because every marker type is a position. A loop is two positions; a memory cue is a position with a label; a grid is one anchor position plus arithmetic. None of them is stored as “the loud part about two-thirds in” — all are times or beat counts measured from the start of a file whose start, length and internal timing MP3Gain leaves exactly as found. This is also why the verification protocol is so short: you are not checking six features, you are checking one property — timing — through whichever feature you know best. If your most-loved cue lands where it always lands, every positional feature on that track is home.

So, the whole answer in one paragraph

MP3Gain cannot move your cue points or beatgrids, because it never re-encodes and never alters timing — it edits a gain value inside each frame and writes undo data to the tags. What it changes is loudness, which is why your DJ software deserves a re-analysis pass afterwards to refresh its loudness-derived data. Keep a backup, test on one crate, verify three tracks by hand, then level the library with confidence. If the remaining question is whether to rely on the software’s own leveling instead, that argument is laid out in rekordbox Auto Gain vs MP3Gain.