Music Technology

Ozone 12 Puts Stem Separation Inside the Mastering Chain Itself

iZotope's new Stem EQ lets producers dial in vocals, drums, bass, and instruments separately on a stereo master file — no original multitrack session required.

Mastering has traditionally been the stage where you work with what you’re given — a finished stereo mix, adjusted as a whole rather than by individual instrument. iZotope’s new Ozone 12 breaks that assumption directly, adding Stem EQ: neural source separation built into the mastering chain itself, letting a mastering engineer apply distinct EQ curves to vocals, drums, bass, and instruments inside a file that was never delivered as separate stems.

Watch: Master Faster With Ozone 12’s Custom Mastering Assistant (iZotope) (YouTube)

What’s actually new in this release

Stem EQ is the headline addition, but Ozone 12 ships three meaningful new modules together: Stem EQ for per-element tonal control without the original multitrack; Bass Control, a machine-learning module tuned specifically to keep low end translating across different playback systems; and Unlimiter, which reconstructs transients and dynamic range from audio that’s already been overcompressed — useful on masters or reference mixes that came in already squashed. On top of those, the Maximizer gets a new IRC 5 limiting algorithm aimed at transparency, and the Master Assistant workflow is now more configurable, letting an engineer set genre targets and LUFS goals and choose which AI-driven modules actually get used rather than accepting a fixed chain.

Why Stem EQ is the one that matters

Neural stem separation on a finished mix isn’t new — tools like RipX and Spleeter have offered it for years as a dedicated, standalone step. What’s new is having it live inside the mastering plugin itself, collapsing what used to be “separate the stems in one tool, EQ them in another, bounce back to stereo” into a single pass. For a mastering engineer handed a client’s stereo bounce with no stems available — a routine situation — that’s the difference between telling the client “I can’t touch the vocal level independently” and actually doing it.

The honest limitation

Neural separation on a fully mixed stereo file is still reconstruction, not the original isolated tracks — artifacts and bleed between elements are a known trade-off of any source-separation tool, Ozone’s included. It’s a genuinely useful rescue tool for masters with no stems, not a replacement for getting real stems from the mix engineer when that’s an option.