Automatic, AI-Powered EQ and Compression for Vocal Tracks

Getting EQ and compression right for vocal tracks is notoriously difficult. It requires keeping track of several faders and knobs simultaneously, including frequency bands, attack and release times, and gain. Hearing all these things at the same time, let alone responding to them, takes years of studio experience and tens of minutes for a single vocal track.

Deprox is a radical new approach to vocal EQ and compression that offers clean and powerful out-of-the-box settings without the need for major tweaks. It has become a drop-in replacement for the first EQ and compression pass in several studios' workflows.

We built Deprox by recording hundreds hours of what we deemed to be "perfect" takes on a variety of microphones. We then introduced subtle imperfections that professional EQ and compression are supposed to correct. Finally, we trained a machine learning model to locate and suppress these imperfections. The result is automatic EQ and compression that rivals some of the most experienced studio ears.

Deprox is intended to be the first plugin in a vocal chain. It doesn't aim to make some of the more radical EQ and compression decisions necessary for a vocal track to sit within a mix. Rather, it gives the best possible baseline for reaching your ideal mix in a fraction of the time.

We came up with the name Deprox because one thing we've consistently heard is that, amongst other things, it fixes the common problem of singers being too close to a mic. A lot of painstaking EQ and compression goes into fixing just this, and Deprox frees you from that grunt work.

Try before you buy! We have a demo page where you can upload your audio snippets for Deproxing.

Examples

From this website, you can download a musical example with Deprox bypassed, with normal mids and with the "enhanced mids" setting, which does more significant processing on the first through fifth formants of a vocalist's spectrum (the 100Hz-2000Hz range).

Interface

Deprox has a single settable parameter, Enhanced Mids, that treats the main vocal formants main aggressively, especially during transients. Most folks prefer this, but it's a bit more computationally expensive. When in doubt, set your DAW's audio buffer length to 512-2048 (the default is usually 128).

The visualization shows where in the frequency spectrum Deprox is acting. It's usual range is +6db to -6db, and it is very active in the 2-5kHz range. In spite of this activity, the result always sounds smooth, as Deprox is working hard to adaptively attenuate proxy frequencies. Otherwise it gets out of the way.

Testimonials.

Deprox makes a vocal track sound less harsh without losing any presence. For ballads and jazz standards, I almost always prefer its result to the original takes.

— Valtteri Tuominen, Sound Engineer, Finnvox studios.

Technical Specifications

Mac OS X 10.13 (High Sierra) – macOS 14 (Sonoma)
64-bit VST3, AAX, and AU, Both Intel and Apple Silicon Macs supported.

Deprox adds on average 0.2 seconds of plugin latency to projects.

As Deprox is often the first plugin in an effect chain, users often commit it (ProTools) or bounce it in place (Logic).

Deprox is a mono plugin. If you need a stereo version, our recommendation is to route your signal through mono busses that each apply Deprox and then mix it back to stereo. This allows you to take advantage of different processing threads, which will result in less CPU usage.