Mix Tools
FREQDIVER
Dynamic spectral unmasking: the main signal steps aside wherever, and whenever, the sidechain occupies the spectrum.
Free 10-day trial. The download is the whole collection, 22 plug-ins in one installer, so you can try every one; start FREQDIVER's trial from inside the plug-in when you are ready. Downloading does not start it, and trying one plug-in does not start the others.
- AU (macOS)
- VST3
- AAX
- macOS Universal
- Windows 64-bit
- v1.0.0
What it does
Make frequencies move out of the way.
FREQDIVER listens to a sidechain and turns the main signal down wherever, and whenever, the sidechain occupies the spectrum, by as much as the sidechain is loud there. Put it on the part that has to make room, feed it the part that needs the room, and the first steps aside only where the second actually is.
Per-bin spectral duckers conflate two different problems. Deciding what the cut should be is a detection problem, solved by 29 band-passes a third of an octave wide reading the sidechain, with no FFT and no latency anywhere in the cut's path. Realising that cut on audio is a filter problem, solved by a swappable renderer. Separating them is why the same cut can be applied with zero latency or in linear phase without the two disagreeing about anything.
Running with no sidechain is legal and does nothing: the cut releases to zero and stays there. A frozen cut on a silent sidechain is the worst possible failure, because it sounds like the plugin broke the track.
Key features
Cut where the sidechain is loud
Each band's cut in dB is proportional to the sidechain's amplitude there. Power is smoothed across neighbouring frequencies before the root is taken, so a single tone makes a smooth bell and a dense sidechain a broad, even cut. At the defaults, a drum bus under a strummed guitar loses 8.3 dB on average while the guitar plays.
No FFT and no latency in the cut's path
Every stage that decides the cut is causal and IIR, each third-octave band with its own attack and release follower. The cut reacts in about 1–3 ms above 1 kHz, 12 ms at 250 Hz and 31 ms at 100 Hz, where a third-octave filter genuinely needs the time. Three STFT layers still run, but only to draw the analyser and to feed Learn.
Dive sets how far the cut spreads
0 % ducks the whole spectrum as one; 50 % is a broad cut around the sidechain's frequencies; 100 % follows the sidechain's spectrum in fine detail. The depth at the sidechain's own frequencies does not change, only how far the cut reaches from them.
Moving notches, measured out
What can sound like phasing in a spectral cut is not phase but narrow notches moving from frame to frame. Power smoothed over octaves has almost none of that fine structure: 1.08 dB of it in the cut, and a worst-case group delay through the whole plugin of about 2 ms.
Two renderers, one cut
Zero Latency uses 40 Cam-spaced TPT bells plus two end shelves, driven by an interaction-compensated least-squares solve and gliding their coefficients per sample: minimum phase, no pre-echo, no latency. Linear Phase uses weighted overlap-add, one window of latency, for mastering and parallel work.
Boundaries you can grab
The sidechain's low and high cut, in cyan, and the reduction range, in amber, are bars on the analyser: drag one, shift-drag a pair together, double-click for its default. The grab zone is sized for the hand rather than the drawing, so the bars are as easy to catch in a laptop-sized window as in a large one. Eight Bias points decide how much each region may be ducked, and Learn can write the sidechain's averaged shape straight into them.
Standard sidechain routing
Declared as a standard aux input bus, so an AU host offers it in its own sidechain menu and VST3 hosts see a second input bus. Mono or stereo main, with the sidechain mono, stereo, or disabled.
The interface
How it works
The signal path, as the plugin’s own documentation states it.
sidechain ─► conditioner ─┬─► 29 ⅓-octave band-passes ─► proportional cut ─► spread, bias, range ─┐
└─► 3× STFT layers ─► 48 ERB bands ─► analyser, Learn │ G[b] dB
main ─────► lookahead delay ─────────────────► renderer ◄──────────────────────────────────────────┘
│
dry ──────► latency delay ────────────► mix ◄─────┘ ─► output ─► bypassAudio
Audio demos for FREQDIVER are not published yet.
We would rather show you nothing than show you a demo that was not recorded through this plugin.
Measured
FreqDiverTests (61) and FreqDiverPluginTests (34), Tools/Profile and Tools/FDRender · Docs/DSP-Notes · 48 kHz
- Test count
- 95 tests, all ten audio-quality criteria green
- auval · pluginval at strictness 10
- clean, on the release bundles
- CPU: Normal, zero latency · linear phase
- 3.3 % · 1.3 % of one core
- Mean cut at the defaults, drum bus under a strummed guitar
- 8.3 dB while the guitar plays
- Fine structure in the cut · worst-case group delay
- 1.08 dB · 1.5–2.1 ms
- Renderer accuracy at 3 / 12 / 24 / 48 dB of reduction
- 0.28 / 0.76 / 1.28 / 3.4 dB
- Attack floor: above 1 kHz · 250 Hz · 100 Hz · 60 Hz
- 1–3 ms · 12 ms · 31 ms · 51 ms
- Null tests: Mix 0 % against the delayed input · Delta against wet minus dry
- −240 dB · −240 dB
Figures are produced by this plugin’s own test suite and benchmark, on the hardware and sample rate named above. They are not estimates.
Controls
MAIN
- AMOUNT
- 0–100 %, default 35. Overall intensity, acting squared. At 35 % a drum bus under a strummed guitar lost 8.3 dB on average while the guitar played.
- DIVE
- 0–100 %, default 50. How far the cut spreads from the sidechain's frequencies, not how deep it goes there.
- ATTACK · RELEASE
- 0.1–500 ms and 5 ms–3 s, default 8 and 180 ms. The band followers, displayed as time to 90 %.
- MIX
- Dry/wet, latency aligned.
TOP BAR
- Presets · SAVE
- Step through or choose the factory and user presets; user presets are readable XML.
- DELTA · SC LISTEN
- Hear exactly what is being removed, or hear the sidechain. In the top bar rather than behind Advanced.
- A · B · COPY
- Two complete settings to compare, and copy between.
- UNDO · REDO
- Across every parameter.
- ADVANCED
- Opens the drawer below. The editor grows; nothing above it moves.
DIVERGENCE & SENSITIVITY
- MAX REDUCTION
- 0–48 dB, default 18. A soft ceiling on the cut.
- SMOOTHNESS
- Cut slope, 9 → 2 dB per Cam.
- GRAVITY
- Makes strong shadows reach further: signal-dependent, unlike Smoothness.
- SENSITIVITY
- ±30 dB of gain on the sidechain before detection: +10 dB behaves exactly like a sidechain sent 10 dB hotter.
- SMART
- 0–100 %, default 30. How much of the cut's dependence on the sidechain's overall level to remove: at 0 % a louder sidechain cuts deeper, at 100 % only its spectrum counts.
- SC LOW CUT · SC HIGH CUT
- 20 Hz–5 kHz and 500 Hz–20 kHz on the sidechain. The same two settings as the cyan bars on the analyser.
- STEREO LINK
- 0–100 %. How far the two channels share one cut, taken from the deeper of the two so the image cannot wander.
- OUTPUT
- ±24 dB.
- LOOKAHEAD
- 0–20 ms. Latency-affecting, so deliberately not automatable.
PROCESS & LEARNING
- PROCESS
- Stereo, Mid, Side, Mid + Side, Left or Right. Mid carves the centre and leaves wide guitars and reverb alone.
- DETECTION
- Adaptive (default), Peak, which catches transients whole, or RMS, which follows power. Scaled to cut the same on steady material at the default times, so the menu changes the character of the response, not its depth.
- RENDER · QUALITY
- Zero Latency or Linear Phase; Eco, Normal, High or Ultra, which sets the analyser's resolution and how often the cut is computed. Neither is automatable.
- AUTO GAIN
- Off by default. Restores the broadband level the cut removed, computed from the cut and the main signal's own spectrum, so an A/B is level-fair.
- BIAS
- Switches on the eight draggable points on the analyser that decide how much each region may be ducked.
- FREEZE · LEARN · USE LEARNED
- Hold the current cut, capture the sidechain's averaged shape, and write it into the Bias curve.
ON THE ANALYSER
- Range bars
- Two amber boundaries restricting the cut to a frequency range. Drag, shift-drag a pair together, double-click for the default.
Technical details
- Formats, macOS
- AU · VST3 · AAX
- Formats, Windows
- VST3 · AAX
- Operating system
- macOS 10.15 or later · Windows 10 or later, 64-bit
- Architecture, macOS
- Universal: Apple Silicon and Intel
- Architecture, Windows
- 64-bit
- Plugin code
- Fdvr
- Version
- 1.0.0
- Latency
- None with the Zero Latency renderer, beyond any LOOKAHEAD you set (0–20 ms). Linear Phase adds one window, 1,024 samples or 21.3 ms at 48 kHz, and gives up to half of it to LOOKAHEAD at no extra cost. Shown in the footer and reported exactly.
- CPU
- 3.3 % of one core at 48 kHz, Normal quality and zero latency; 1.3 % in linear phase
- Factory presets
- 12
Vocal Space — Gentle, Vocal Space — Strong, Kick vs Bass, Snare vs Guitars, Lead Synth Focus, Dialogue Clarity, Reverb Ducking, Delay Ducking, Mid Vocal Carve, Mastering Subtle, Broad Separation and Tight Separation. User presets are saved alongside them as readable XML.
Worth knowing
- Attack cannot outrun the filter bank at the bottom. A third-octave band-pass needs time to ring up, and more the lower it sits: however fast Attack is set, the cut takes about 1–3 ms above 1 kHz, 12 ms at 250 Hz, 31 ms at 100 Hz and 51 ms at 60 Hz. Above those floors the knobs hold.
- Renderer accuracy is depth-dependent: 0.28 dB at 3 dB of reduction, 3.4 dB at 48 dB. This is a measured limit of the filter family, not a tuning issue. The default MAX REDUCTION of 18 dB sits between the 0.76 dB measured at 12 dB and 1.28 dB at 24.
- The proportional cut was fitted and checked on one pair of tracks, a drum bus under a strummed guitar, and on synthetic probes. A vocal over a full mix and a sparse arrangement have not yet been measured.
- The linear-phase renderer has had far less attention than the zero-latency one. It nulls correctly and reports the right latency; its pre-echo has not yet been measured.
- The factory presets were voiced before the cut was redesigned and have not been re-voiced: they now cut between 1.1 and 2.7 times as deep as when they were set.
Also in these bundles
COMPLETE COLLECTION
Get FREQDIVER as part of COMPLETE COLLECTION, 22 plugins worth 708 for 199.
199
Save 509
