iZotope Ozone 12 Review: Can AI Mastering Replace Your Engineer?

Ozone has been the default “mastering in a box” answer for years now, and every version since the iZotope machine-learning push has asked the same question: how much of this can genuinely be trusted to an algorithm? With Ozone 12, that question has more context than usual — iZotope was acquired by Boris FX in July 2026, after a turbulent stretch under the Native Instruments umbrella. So this isn’t just a review of a plugin. It’s a review of a plugin from a company that just changed owners and is publicly promising “another generation of intelligent audio technology.”

I use Ozone on real deadlines — client EPs, sync work, the odd self-released single where I’m both artist and engineer. So let’s talk about what Ozone 12 actually does in the room, not in the promo video.

First, the elephant in the room

You can’t review Ozone 12 in mid-2026 without addressing the Boris FX acquisition, because it directly shapes whether you should trust the roadmap. The short version of the story: Francisco Partners took a majority stake in Native Instruments in 2021 and rapidly swallowed iZotope, Brainworx, and Plugin Alliance. By November 2025 a joint acquisition by Bridgepoint and Bain Capital had failed, and in January 2026 Native Instruments entered preliminary insolvency proceedings. inMusic picked up Native Instruments and its controlled brands in May 2026, and then in July 2026 Boris FX acquired iZotope, pulling it out of the NI family entirely.

That sounds alarming, but the practical picture is calmer than the headlines. Boris FX has publicly committed to investing in iZotope’s future, stated that plugin development continues without interruption, and confirmed that the teams behind RX, Ozone, Neutron, and Nectar stay in place. iZotope operates as a dedicated unit within Boris FX, the existing engineering, product, and support teams remain intact, and all existing products, licenses, and subscriptions are expected to remain fully functional with uninterrupted support. One real change: iZotope products are no longer available via Plugin Alliance — though that could shift again.

So if you already own Ozone 11, your licenses still work. If you’re buying Ozone 12 today, you’re buying from Boris FX. The 30-plus-year history of that company in pro film and VFX tooling suggests they know how to ship demanding software — but whether they keep serving people who just want to finish a track and not a film is the open question. I’d keep an eye on it rather than panic.

What the AI actually does — and what it doesn’t

Here’s the honest framing. Ozone’s “AI” isn’t a black box that replaces your taste. It’s a reference-matching and diagnostic engine — it listens to your mix, suggests target EQ curves, dynamics, limiting, and stereo width based on a reference track or a genre profile, then hands you a plugin chain you’re free to ignore.

That’s the crucial bit. The Master Assistant gives you a starting point, not a finished master. In practice, I load it, let it propose a chain, then immediately start gutting things:

The corrective EQ suggestions are usually the most useful part. If you’ve been mixing in a bad room, Ozone often catches a low-mid buildup or a top-end tilt you’ve stopped hearing.
Multiband dynamics gets aggressive out of the box. I almost always back it off. AI tames dynamics because it thinks in terms of “safe,” and safe masters sound flat.
The limiter is doing more heavy lifting than any “intelligent” module. Nothing else in the chain matters if the limiter is working against your transients.
Stereo width suggestions lean wide. Great for pop, disastrous for anything that needs a mono-compatible, focused low end.

The point: Ozone 12’s AI is a second pair of ears that never gets tired. It is not a second pair of ears that knows your artist intent. That distinction is where a lot of the “can it replace your engineer” debate goes wrong.

Loudness, streaming, and the -14 LUFS trap

This is where AI mastering plugins earn or lose their keep. Streaming platforms normalize loudness, so a huge part of “mastering for streaming” is making sure your master’s integrated loudness and true peak ceiling are sane for the target platform — and crucially, that your dynamic balance still translates after the platform’s own normalization knee.

Ozone’s loudness targeting is genuinely helpful here. You point it at a target, and it does the boring math so you’re not guessing. But a number is not a master. I’ve heard masters that hit the streaming target perfectly and still sound lifeless, because the transients got squashed to get there. The AI will happily help you do that if that’s what you ask for. Garbage in, normalized garbage out.

The right way to use the loudness features: set your ceiling, then A/B against your uncompressed mix and ask whether you’ve gained weight or just lost punch. If it’s the latter, pull the limiter back even if you land a couple of LU below target. Nobody’s ears ever complained that a track was 1.5 LU under a streaming target. Plenty have complained a master sounded crushed.

The real workflow advantage

Strip away the AI marketing and Ozone 12 is still a deeply competent single-window mastering chain: EQ, dynamics, exciter, imager, maximizer, all in one plugin with assistive presets. For anyone self-mastering — especially on a laptop, on a deadline, without a treated room — the value is the time it saves getting to a defensible starting point.

That starting point matters most when you’re the one who mixed the track. You have “mix ear” — you’re so used to your balances that you can’t hear the master objectively. Ozone breaks that loop, because the suggestions come from an analysis rather than from your memory of what you intended. That’s genuinely useful. It’s the same reason producers use reference tracks.

What it can’t do is make the call that only a human makes: knowing that the artist wants the vocal slightly buried, that the bridge should feel unmastered, that the whole thing should sound worse in a specific way because that’s the aesthetic. AI optimizes toward “good.” Good is not the same as right. Any engineer who’s worth the fee is paid for that judgment call, not for the EQ curve.

So, does it replace your engineer?

For demos, for quick client previews, for anything that isn’t the final deliverable — yes, absolutely, and better than most self-mastering attempts. For radio-ready singles or anything where your reputation rides on it? No, and I’d be suspicious of anyone claiming otherwise.

Here’s the nuance that gets lost: Ozone 12 doesn’t replace your engineer. It raises the floor of what “no engineer” can sound like. That’s a different and more honest claim. The bottom of the market gets squeezed — the “I’ll run it through a preset for $50” tier of mastering is genuinely under threat. But the top of the market — a pair of trained ears and a lifetime of taste, in a room you trust — is not. That engineer is still hearing things no analysis window will flag.