iZotope Velvet

Could this ‘smart’ plug‑in help you massage vocal and dialogue recordings into better shape?

I’ve been impressed with iZotope’s new Catalyst range of plug‑ins. Essentially, they add a dollop of ‘smart’ (ie. machine‑learning assistance) to one or more of the sophisticated processing options found in the likes of Ozone, Nectar and Neutron, and make it available in a more streamlined form at an accessible price. From a user point of view, it’s an approach that has a lot to commend it. The latest addition to this range is Velvet, a smart de‑esser, dynamic tonal balancer and noise (mouth sounds, pops, clicks) remover, which is intended primarily (if not exclusively) for processing vocals or dialogue.

Three‑pronged Attack

At the top of the GUI, the global control strip features buttons for delta auditioning (the difference between the dry and processed signal) and bypass, as well as preset access and a Mix slider. The processing is applied in real time, and my Cubase system registered around 100ms of latency, so it will be fine for mixing but is perhaps best not used while tracking There are three main panels: Sibilance, Tonal and De‑Click. The incoming audio signal is split before being passed through the Sibilance and Tonal modules. Processing is done independently (in parallel) in these two modules before the recombined signal passes through De‑Click.

At the heart of the Sibilance module lies a sophisticated multiband dynamic EQ. You can configure this manually to using up to six dynamic (or static if you prefer) bands, each with frequency, gain and Q controls, as well as the option to select bell or low‑/high‑shelf filter types, or to apply upward or downward compression. But a Learn button offers intelligent assistance: press this and Velvet analyses the incoming signal and makes a best guess at the Amount (of gain reduction) and Threshold (the point at which processing is triggered) in each band, which you can then adjust to taste. The module’s Solo button makes it easy to judge precisely what this processing is doing.

The Tonal module is similar in principle to the Clarity or Sculptor modules found in iZotope’s Ozone and Neutron, in that it uses dynamic EQ to adjust the tonal balance. Its Lift control lets you boost frequencies that might be too low in level, while Tame does the opposite. The Lows, Mids and Highs sliders control how intensively these dynamic changes are applied to three broad frequency bands. And, to give Tonal a nudge in the right direction, there’s a selection of vocal and dialogue targets (and some Instrument and Cymbal options) to identify the source material type.

The De‑Click module’s Amount control is joined by Freq Skew, which focuses the detection more or less on the highs/lows if required. As with the other modules, there’s a Solo button should you need it.

The presets make it easy to find a sensible starting point but, importantly, you can then adjust the settings to taste.

Velvety Smooth?

Velvet’s GUI is compact and intuitive, and I found this plug‑in very effective on a range of spoken and sung vocal recordings. The presets make it easy to find a sensible starting point but, importantly, you can then adjust the settings to taste. It’s usually a simple matter to tweak the main controls, perhaps using the Delta or Solo buttons to hear precisely what your processing is achieving and to listen out for artefacts. While cleaning up vocal recordings and achieving a more balanced sound is simple, I also found Velvet great at taming splashy cymbals, both in a multitrack drum recording and, albeit with a little more tweaking, within a stereo drum loop.

Yes, you could probably achieve similar results with three separate plug‑ins, not least with iZotope’s ‘full fat’ tools. But why bring three plug‑ins into the project where one will do? Velvet offers a very neat combination of sibilance reduction, tonal balancing and de‑clicking, and the real‑time insert plug‑in format is very convenient. Its price is modest too, so it’s convenient, compact and cost‑effective. Whether you’re working with vocals or dialogue, the 10‑day free trial is well worth a spin.