Mastering an EP or album of AI tracks so every song sits at the same level

Why AI tracks vary more than you expect

Two Suno generations from similar prompts can differ by 3 dB in loudness and by several dB in low-end energy. In our 20-track measurements the integrated loudness ranged from about −17 to −12 LUFS across the set. Each track was fine on its own; together they sounded like a playlist from five different producers.

Target the same number, then listen

1. Master every track to the same integrated loudness and ceiling (for streaming, −14 LUFS and −1 dBTP). This removes the big level jumps automatically.
2. Play the tracks back to back and listen to the transitions. A ballad at −14 next to a club track at −14 can still feel wrong, because integrated loudness averages the whole song.
3. Adjust by ear, within ±1 dB. Quieter songs are allowed to be a little quieter; that is what album mode on streaming preserves.
4. Check tonal balance across the set. The same style reference for every track keeps the low end and the top end consistent, which does more for 'sounds like one record' than the level does.

Use the same chain, not the same settings

Copying one track's EQ and limiter settings onto the next is the most common album-mastering mistake. Each track has different starting measurements; what should be the same is the target and the reference, not the knob positions. An engine that measures each file and re-designs the chain per track gets consistency for free, because every track is pushed toward the same reference curve.

Order and gaps

Decide the running order before you master, so the transitions you listen to are the real ones. Keep gaps consistent (1.5–2 s is typical) and trim the silence in the files themselves rather than relying on the distributor.