We're getting into the age of intelligent plug-ins. They are doing so much that it's hard to keep track of how all these little changes stack up when in use. I recently began trying out NoiseWorks DynAssist on John McEntire's [Tape Op #23] recommendation, as he said it helped with cleaning up and leveling out vocal tracks. When NoiseWorks recently released VoiceAssist, I dropped John a note and asked if he was using it, and I got back, "Yeah, man! It’s f**king INSANE!!! There is no going back." This plug-in works in ARA* mode, Audio Random Access [see tutorial] , where clips or tracks are selected and processed in tandem with a DAW, and it currently works with Pro Tools 2024.10+, PreSonus Studio One, Reaper, and Steinberg's Cubase and Nuendo. A standalone app comes with this plug-in, which can be a quicker way to work at times. That’s also a big perk for anyone using this for video editing. It has a non ARA compatible DAW transfer mode that will “record” and analyze the clip as you play it back and then perform its changes.
VoiceAssist is a multifaceted tool for prepping vocal tracks, with toggle-able modules for Clean, Lowend, Gate, Dynamics, Breath, and Sibilance that address common issues with both sung and spoken parts – these can be quickly accessed with simple key commands like S for Sibilance. Opening the app, you'll see a waveform view as the program analyzes the audio (note the progress status in the bottom right panel). There are toggle buttons for the modules mentioned above, and clicking on their names will bring in a different view with functions that can be applied.
It's a lot to look at and navigate, and a lot to even think about, but I found myself letting VoiceAssist do its thing and then bypassing it to see what I got. Even the standard default preset sounded cleaner and more dynamically balanced. For spoken word, Breath reduction alone could be a lifesaver. The Sibilance control module is one of the best de-essers I've ever heard; I didn’t notice it working, yet S sounds were softened up naturally. The Clean module is the real mystery section. It performs AI-based denoising, dereverberation, and frequency repair. Initially it was the least controllable section of VoiceAssist, but a recent update added spectral editing that allows for some fine tuning, though the image is very low res and the controls are clunky. The Recover Highs toggle button and a frequency selector are handy for undoing bad mic choices or adding clarity.
The one thing I couldn't really get VoiceAssist to address perfectly was low-end plosives. Clean removes some but not all, even at the “infinite reduction” setting. I frequently use the De-plosive module in iZotope RX's app to identify and remove unwanted popped Ps and such, but here I could only apply a Pitch-based Lowcut that sometimes felt inactive, or an adjustable Lowcut that would thin out the sound if set at too high a frequency. Then again, Lowcut was handy for digging the mud out of vocals where proximity effect was too thick. VoiceAssist currently doesn't come with any presets, but they tell me they will in future updates. I wonder if presets would help as starting points; I plan to build a number of my own, named "Eating the SM7" and such. Another vocal sound I always have problems with is spikey transients from consonants like hard C or G sounds. In RX, I'll be De-clicking those specific spots or even diving into Spectral Repair. VoiceAssist doesn't address sounds like this, and rightly so, as smoothing all these transients out could adversely affect vocal clarity. This is just something I want any user to be aware of and keep an ear out for.
I should point out that VoiceAssist will run the preset and do all the work, but it's easy to go in and make specific adjustments along the timeline. Say there's one gulping breath you don’t want to hear; it can be pulled down by zooming in and simply grabbing the bar at the top of that breath (it shows the amount of dB reduction applied). This applies to Sibilance as well, so hiding a big S or undoing an over-reduction is easy. As VoiceAssist breaks the processing into blocks for analysis, you can change parameters during playback, and it will quickly adjust the track you are hearing and then analyze and change the surrounding blocks of audio. For all the work going on, it seems pretty efficient. The two different Apple Macs I used handled the workload with no issues. That said, NoiseWorks told me to remind you to check the app’s system requirements before installing.
I spend a lot of my mixing time prepping tracks, and a majority of that work is on vocals. My short article in Tape Op #170, "Top 12 Vocal Recording Mistakes," covered a lot of the issues I find with tracks I get sent to mix. VoiceAssist is a quick way to address many of these issues, and I found that the worse the source track was, the more noticeable the results. There still may be other issues to address, but VoiceAssist can be a big help and a time saver for busy mix engineers and in other post-production work.
_________________________________________________________________________________________________
*ARA = Audio Random Access
What is ARA? I think it's an important concept to understand in digital audio production. Basically, think of this as two different computer applications working in tandem to supply audio files to the DAW (digital audio workstation) that have been externally modified yet can be continuously updated. Think of it this way: You tuned a vocal track outside of your DAW and then rendered it. Then you hear one spot where the tuning app went buggy on a note, and it's not cool. This is going to require a bit of workaround. If I’m using Celemony's Melodyne via ARA in my Avid Pro Tools session, I can quickly jump over to the ARA window and fix the issue. No time spent sending a track to an outside app, no weird processing, and no rendering (unless you want to totally lock it in).
At this point (as far as we know), the DAWs that feature ARA implementation are Avid Pro Tools, Fender Studio Pro, Apple Logic Pro X, Universal Audio LUNA, Cakewalk by BandLab, Reaper, Acoustica Mixcraft, Tracktion Waveform, Boris FX Samplitude, and Steinberg's Cubase and Nuendo.
Now, 2018's ARA2 is likely the version you are using. It introduced more undo capabilities within your main DAW, better multitrack support, and tighter syncing. Synchronization is something to seriously consider though, as I've never seen a warning panel telling me my ARA tracks are not reintroduced back into my session fast enough. Think about the phase issues that could happen with two applications working in tandem. But man, the benefits so far are amazing…
-LC