Acon Digital Extract:Dialogue 2

Need to clean up a dialogue recording without too much hassle? Then Acon’s latest tool might be worth checking out...

Acon have long been pioneers in the field of machine‑learning‑based audio processing, and I’ve been a user of various of their plug‑ins for almost as long. They might not always be the first tools I reach for, but sometimes they are, and they’ve often dug me out of a hole when nothing else seemed able to. One tool that I use frequently, and which has rescued more than one podcast project for me, is their Extract:Dialogue plug‑in. So, on discovering that this had been treated to a significant update, I was particularly keen to check it out.

New Features

Even a cursory look at the drag‑to‑resize GUI tells you that lots has chanced since v1. Not only does it look much more slick and modern generally, but there are also many more controls. Notably, the left of the GUI now hosts controls for three separate ‘channels’, where there were previously just a couple of global controls. Each channel has a +10 to ‑40 dB gain fader and numeric readout/direct‑entry box, along with solo and mute buttons, and each is dedicated to a different signal type: Voice, Reverb or Noise. (No prizes for guessing that their content is identified behind the scenes using neural networks!) This sort of three‑fader interface seems to be becoming ubiquitous for such tools: it’s similar to the arrangement used in iZotope RX’s Dialogue Isolate plug‑in and Accentize’s recently launched dxSplit, for example. That’s hardly surprising, since it’s super‑intuitive: essentially, you just drag the faders up/down for more or less voice, reverb (room tone) or noise (non‑voice sounds) in the recombined signal that comes out of the plug‑in.

Acon managed to pack a lot into a simple interface: note the pink button at the bottom of this screen, which tells you the Reverb channel is in view, and the orange EQ button that tells you the display is for the EQ, rather than the sensitivity.

To the right of all this, that same ‘cursory look’ suggests a similar three‑band frequency‑sensitivity display as in version 1, but actually it offers significantly enhanced functionality and options. Each of the three ‘channels’ is given its own tabbed window on the main sensitivity display, selectable either by clicking on the logo above that channel’s fader, or a button below the display. The plot defaults to showing sensitivity for the currently...