When they’re not busy slurrying humanity down to the biofuel necessary for their own continued existence, our future robot overlords will need music and movies to fill the downtime. With no one left to produce said media, though, that job will fall to the machines too – machines that will undoubtedly descend directly from terrifyingly good artificially intelligent plugins like these…

Accentize Chameleon 2

A dream plugin for post production in particular, Accentize’s cerebrally awakened reverb effect draws on AI to make matching the ambience of replacement dialogue to that of the original location recording a snap. Simply play a few seconds of the location audio into Chameleon and a neural network generates an impulse response-based preset to recreate its spatialisation. The resulting patch might sound just right straight away or require a bit of manual tweaking using the provided controls, which include pre-delay, decay and filter parameters, as well as dry/wet mix (dialled in automatically as part of the preset generation process). The impulse response itself can be exported for use in other reverbs, too, and there’s nothing stopping you from playing dry sounds of all kinds into the plugin in the name of creative experimentation.

Also useful for music production, when you need to overdub instrumental parts onto tracks recorded elsewhere.

Hush Audio Hush

There’s no shortage of brilliant audio restoration plugins and applications out there for the dialogue producer looking to eliminate background noise and reverberation from voice tracks, but Hush Audio’s contribution to the market is notable in that it works its signal-scouring magic entirely locally (rather than via upload to the cloud, as some others do) and directly exploits the much-vaunted Neural Engine integrated into all Apple Silicon-powered Macs for much faster processing on those computers than their Intel forebears.

Everything is done by the AI, with the workflow involving no user input beyond dragging the files to be processed into the interface, setting the output file format, sample rate and bit depth (if different from the source), and balancing the dry/wet mix. And as well as sustained noise and reverb, Hush also takes care of unwanted transient sounds such as birds, door slams, car horns and the like. It’s currently a stand-alone application, but an Audio Units version is on the way, we’re told.

Baby Audio TAIP

Baby Audio’s handsome little plugin is the first analogue tape simulation effect to build artificial intelligence into its technological foundations, and the end product is truly impressive. On the surface, TAIP looks like any other tape sim, with knobs and sliders for adjusting overdrive, frequency-shaping and various electromagnetic/mechanical contributors; but behind the scenes, an algorithm highly trained in the differences between dry and processed signals dictates the actual sound of the virtual tape and the machine moving it. The authenticity of the thing is quite amazing, and the Noise, Wear (aging, wow/flutter) and Glue (compression) parameters open up plenty of creative saturation possibilities.

iZotope Neutron 4

Few would reasonably argue that the meat-based mix engineer is in genuine danger of redundancy any time soon, but when the audio AI apocalypse does finally arrive, its Skynet analogue will surely be iZotope, who have gone all-in on machine learning with their flagship plugins in the last few years. Neutron’s Mix Assistant feature is among the most promising implementations, spitting out an optimal target ‘tonal balance curve’ and corresponding ‘starting-point’ processing chain built with Neutron’s main modules (EQ, Compressor, Exciter, Gate, etc) after only a few seconds of input analysis. Spectral profiling enables imported reference audio clips to guide the AI and its recommendations, while the Punch, Distort and Width macro controls make pushing the processing towards the target curve as intuitive as their ‘outcome orientated‘ categorisation suggests.

There’s no way you’re just going to be slinging the Neutron Mix Assistant on every channel and calling it a day (yet!), but as a time-saving guide (and educational aid), it’s well worth investigating.

Sonible smart:bundle

Resolutely focused on the deployment of AI for the automation of mix processing since the formation of the company in 2016, Sonible have rightly earned plaudits galore with their ‘smart’ series of plugins, which can be bought individually or as a set. Comprising the largely self-explanatory smart:comp 2, smart:limit, smart:EQ 3 and smart:reverb, smart:bundle lets you relinquish the mix engineering essentials to the whims of the machine, each one setting up the parameters for its titular process to bring out the absolute best in the input signal.

What qualifies as ‘best’ depends on the sonic domain in question, of course: smart:comp 2 and smart:limit aim for transparent compression, ducking and limiting; smart:EQ 3 shoots for a well-balanced frequency response with no awkward resonances; and smart:reverb conjures reverb effects tailored to the specific tonality of the source material. Helping to realise these goals are various innovative technologies, including smart:comp 2’s powerful spectral compression (up to 2000 frequency bands!), smart:EQ 3’s agile cross-channel processing, and smart:reverb’s style-wrangling Reverb Matrix. Every plugin also features a panel of user-accessible parameters for fine tuning and creative manual override.

Deeply adaptable, worryingly brainy and thoroughly effective, smart:bundle feels like the future of mixing.

Acon Digital Extract:Dialogue 1.5

Another AI-powered noise reduction solution, Acon’s acclaimed dialogue clean-up plugin recently saw a significant update that not only gave it a boost in terms of output quality and processing speed, but also dialled down the CPU usage and introduced phase correction for more natural sounding results. Extract:Dialogue’s interface couldn’t be simpler, with only a handful of optional controls to negotiate (Sensitivity, Maximum Attenuation and two filters), while the extensively trained machine learning engine works its magic in the background. And like Hush, all processing is carried out locally, with not a cloud in sight.

Waves Clarity Vx DeReverb

The newest kid on the AI plugin block is a reverb-nuking version of Waves’ established Clarity Vx noise reduction processor that only came out a few weeks ago. Like its sibling, Clarity Vx DeReverb comes in regular and Pro versions, both of which are governed by the same set of neural networks and centre on a big main knob for the immediate melting away of unwanted ambience. The pricier Clarity Vx DeReverb Pro, however, adds multiband operation (six bands), higher maximum reduction, tail smoothing (“to rebuild the room and help dialogue to match the visuals”) and a limiter, for greater flexibility at the expense of simplicity and, er, money.

See and hear both versions in action in Paul Maunder’s video.

Have artificially intelligent plugins wormed their way into your production workflow yet? Tell us all about it in the comments.