The choice of AI-powered dialogue repair plugins has surged over the last few months, many of which we’ve covered on the blog. In this article and video, Paul Maunder puts one recent release to the test on a number of different dialogue recordings, each with its own noise problem, to see how it performs…
You may be familiar with the Hush app for Mac OS developed by Ian Sampson. Hush uses machine learning to remove noise and reverb from spoken audio. This standalone app has been available for a while but now, with the introduction of Hush Pro, the same functionality and processing is now available as an AAX AudioSuite plug-in.
Hush Pro actually consists of two components: Hush Mix and Hush Split. Both plug-ins use the same engine under the hood, but offer two different workflows.
Hush Mix
Hush mix separates dialogue recordings into their Voice, Reverb and Noise components and provides channel faders for each. Up to 36dB of cut can be achieved on the fader, or up to 12dB of boost. Each channel provides a solo button, along with a mute so you can cut reverb or noise out entirely if required. There’s also a Bypass button for each channel, allowing A/B comparisons to be made before you commit to rendering the result. Link controls allow voice, reverb or noise faders to operate together. This can be handy when you want to eradicate both the reverb and noise components of a recording simultaneously.
Hush Split
The Hush Split AudioSuite plug-in presents itself as a simple set of written instructions and a render button. This allows the voice, reverb and noise to be separated and rendered onto their own playlist lanes or audio tracks. The ability to split the audio in this way provides the option to balance out the components using the channel faders in Pro Tools or to extract room tone from a dialogue recording.
In Use
I tried Hush Pro on a number of different dialogue recordings, each of which contained its own noise problem. Hush Pro proved to be an effective way of reducing both reverb and noise in dialogue recordings. There have been quite a number of new machine learning based noise reduction plug-ins released over the last few months and the bar has been set higher than ever in terms of what can now be achieved. Fortunately, Hush Pro compares favourably, doing a great job at separating dialogue from room reflections, ambient noise and transient sounds. No noise reduction plug-in is perfect and I did find some cases in which Hush Pro wasn’t able to fully remove the noise from a voice recording. In the majority of cases though, it not only completely removed the noise and reverb but also left the audio more intact than many other products on the market. The lack of audible artefacts is one of the most notable differentiators of Hush Pro.
To see and hear Hush Pro in action, check out the video below.