Guides

Why AI songs sound narrow, and how to widen without ruining mono

Updated 2026-09-06

Put a commercial pop mix on a correlation meter and the side channel sits 6 to 10 dB below the mid. Put a Suno or Udio export on the same meter and the side sits 15 to 20 dB below. That is the difference between a record that fills the space between your ears and one that sits in the middle of your head.

Why generators make narrow mixes

Most models generate something close to a mono signal and add a little decorrelated width late in the process. Reverb tails and doubled parts, the things that make a human mix wide, are produced as part of the same texture rather than placed.

Widening that works

• Only above 200 Hz. Below that, width turns into phase problems and a bass that disappears in mono.
• Raise the side signal by 2 to 4 dB, not more. Past that, the center collapses and vocals lose focus.
• Add the width per band. A little more in the highs, where reverb lives, and almost none in the low mids.
• Check mono. Sum the channels and listen for anything that vanishes. If the vocal drops, you went too far.

Bass stays mono

Club systems sum the low end to one channel and vinyl cutters cannot handle stereo bass. Keep everything below roughly 100 Hz in the center, always. A widener that touches the bass is a widener that will get a track pulled from a DJ set.

Burnish measures side versus mid on your file, widens above 200 Hz toward the genre target, keeps the bass mono below the profile's crossover, and prints the before and after width in dB on the report.