Tone
Bass, mid, and treble shaping with mode-aware shelf
and peak placement. Mid sits at 2 kHz in Vocal mode
for speech intelligibility, 1 kHz in General mode
for warmth.
Speech Clarity
Four independent bands - de-esser, plosive control,
breath reduction, and click/mouth-noise repair - for
targeted speech-artifact cleanup without touching
the rest of the signal.
Denoiser
Broadband denoise for hiss, fan noise, and steady
background beds. Runs a classic spectral noise
floor estimate rather than a neural model, so CPU
use stays low enough for dozens of tracks at once.
GR meter shows active noise reduction in real
time; use it to confirm the reduction amount
before trusting a strong setting by ear alone.
Deverb
Room-tail reduction tuned to act on late ambience
more intentionally instead of flattening the whole
source.
Finish
Final-stage polish combining a Recovery lift
stage with true-peak-aware limiting, so it can
add perceived loudness back after corrective
processing without letting inter-sample peaks
clip downstream. Meant to be the last plugin in
the chain, after tone and dynamics are settled.
OptoComp
LA-2A-style opto levelling with programme-dependent
gain reduction. Pro mode unlocks Behavior (Compress
or Limit) and Stereo Link for engineer-facing
control.
Proximity
Source-aware close-mic modeling for intimacy, body,
and air, with vocal-focus adaptive pattern factor.
Subtract
Learn a repeatable noise profile, then subtract it.
Now warns when a profile is older than 20 minutes
and the noise floor may have drifted.
Leveler
Adaptive levelling with two behaviours on one
control set: a true vocal rider that freezes its
fader in pauses and can hold a vocal a set amount
above sidechained music, and a mix leveler with
realtime, smart, and timeline-locked offline
analysis. Target and Depth trim where the ride
sits and how far it reaches; the level trace shows
the fader and its target as it works.
DeepFilterNet
ML-backed denoise for heavier or more complex
interference than a spectral denoiser can track
cleanly - traffic, chatter, non-stationary beds.
The Guard control blends the model's output back
toward the dry signal when the network starts
producing artifacts, trading some noise removal
for a more natural result on hard material.
Choose from DeepFilterNet 3, DeepFilterNet 2, or
the lighter RNNoise model; inference runs on a
dedicated per-channel thread, never the audio
thread, so the model can't glitch the signal
path.
Rebalance
Separates a stereo mix into Vocals, Drums, Bass,
Guitar, and Other on the fly and lets you push
each family up or down, no stems required. A
confidence signal scales how hard each move is
allowed to push, so an uncertain separation eases
off automatically instead of smearing the mix.
ToneRefine
Automatic mud, harshness, and roughness correction
using detection-driven biquad processing. Targeted
subtractive refinement without manual EQ.
Repair
All-in-one guided repair built on the same
embedded DSP as Denoiser, Deverb, and Speech
Clarity. Analyses the audio and suggests settings
per stage - Noise, Click, Clarity, Reverb - rather
than one global amount, and a scale control on
each stage lets you dial an individual suggestion
back without redoing the analysis.
Studio Analyser
Chain-wide dry-vs-wet spectrum view. Insert it
last in the VX chain, then use the Full Chain
button to compare chain input against final
output, or click any stage in the left rail to
inspect only that processor's own dry-vs-wet
spectrum. It follows only the active upstream
path feeding the instance you are inspecting, so
it stays correct even when other tracks route
through the same plugin chain.
Tune
Confidence-gated vocal pitch correction: instead
of snapping every note to the nearest scale
degree, it checks whether a deviation looks like
an error or an intentional bend/vibrato before
acting, so expressive performance detail survives
correction. The live sung-vs-tuned trace lets you
see exactly what it changed, note by note, rather
than trusting the setting by ear. Key and Scale
can be set manually, detected automatically from
the vocal, or - uniquely - learned from an
instrumental routed into the sidechain input:
route the backing track in the way you'd route a
sidechain to a compressor, and VXTune reads the
actual chord content instead of guessing the key
from one voice at a time. Most tuners either need
a separate key-detector plugin and a cross-plugin
key broadcast, or ask you to enter the key by
hand - VXTune just needs a send.
Width
A target-seeking stereo image engine: instead of
applying a fixed widening curve, VXWidth measures
the source's actual current stereo width and
solves for the existing-Side gain needed to reach
the requested target, so a narrow source gets
more expansion than one that's already wide at
the same knob setting. Double adds a synthetic
doubled performance (with an always-on subtle
micro-pitch stage), Tightness controls how
closely that generated performance follows the
original, and Focus places its spectral content
(Body, Full, or Air) - none of those three can
move Width's own output, even when Double is
active. Blend then balances the whole result
against the original signal: 50% is the normal
VXWidth sound, left leans toward the untouched
input, right pushes further in the same
direction the effect already moved it. A
harmonic/transient-aware decorrelation stage and
a full mono/phase-risk guardrail (per-band
coherence, downmix spectral deviation, centre-image
displacement) run underneath throughout, so
widening and doubling stay mono-safe on real
material.
