Online AI mastering for songs: when and how

Master after the split; unify loudness; do not expect clipped audio to revive

When to master online

Online AI mastering for songs is a loudness and balance pass, not a replacement for tracking. On 魔镜音频 it usually sits after vocal separation: split the lead, confirm backing layers, then one-click master if you need a listen-ready bounce. Skip it if you are taking stems into a DAW.

Open www.1momirr.com, upload, finish the split (when you need it), enable mastering, preview on headphones and a phone, then download WAV and lyrics.

Upload-to-download checklist

Check the source for clipped peaks before you upload. If you need a lead/accompaniment split, do that first and A/B the chorus. Enable one-click mastering only after the bed still sounds usable. Preview on two devices. Download WAV; avoid stacking lossy exports.

Five failure points: already-clipped source; backing vocals squashed after the loudness push; chasing LUFS while ignoring sibilance; refreshing during the job; checking only one speaker.

Honest limits

Mastering will not restore missing peaks, un-smear a long reverb, or match a custom analog chain. It is for fast unification after a split, or a demo bounce. Fine automation and taste still need a person.

If you skip separation and only want a louder mix, that is allowed—but say so. Related: Vocal separation, How to separate vocals online, and the Chinese mastering boundary list.

Should I master before or after vocal separation?

After, if you care about backing-vocal depth. Mastering a bed that already lost harmonies just makes the hole louder.

Does this replace an engineer?

No. It compresses install-and-tweak into upload–process–preview–download. Use it to verify quickly; keep humans for the last taste calls.