Spectral Agent

{See video at top of page}

Spectral Agent - Auto EQ/Unmasking Plug-In

Priority-Driven Dynamic Spectral Resolution:

• Drop Spectral Agent on any/all of your tracks (up to 64).
• Create your network of Master and Yielding interactions.
• Let the Auto EQ unmask and bring clarity to your mix.
• No sidechaining necessary - Spectral Agent instances pass data automatically.

Windows and MacOS-compatible (VST3 and AU)

Demo version available for free here: https://immanentaudio.gumroad.com/l/cveia

The Challenge of Dense Arrangements In modern mixing, tracks inevitably compete for the same frequency bandwidth. Static EQ can carve out space to resolve these collisions, but those cuts are permanent—leaving unnecessary hollows in a track even when the competing instrument isn't playing.

The Solution: A Priority Ecosystem Spectral Agent replaces static cuts with a dynamic, project-wide frequency ecosystem. By deploying instances across your mix, you define a clear acoustic hierarchy. A lock-free internal registry allows tracks to communicate and share spectral data in real time. No sidechaining involved.

Instead of guessing where masking occurs, a Master instance broadcasts its dominant spectral peaks. A Yielding instance reads these peaks and deploys highly targeted, transient-responsive notch filters exactly where—and only when—a collision happens. When the Master stops playing, the notches release, preserving the natural body of the Yielding track.

Key Architecture & Features

• Project-Wide Priority Interactions Spectral Agent scales across your session with support for up to 64 networked instances. A single track can simultaneously act as a Master track in one relationship, and a Yielding track in another, allowing you to build a comprehensive hierarchy without complex sidechain routing.
• Intelligent Slot Allocation The engine utilizes seven general-purpose hysteresis-controlled notch slots, plus one dedicated Bass slot. When multiple tracks compete for space, Spectral Agent uses a Weighted Priority Algorithm to ensure the most critical elements win the frequency allocation.
• Acoustically-Tuned Envelopes Attack and release behaviors are adaptive. They scale automatically across the frequency spectrum—preventing artifacts and retaining a natural feel whether you're ducking below 100Hz or above 10kHz.
• Precision Visual Feedback An 80-band log-scale analyzer renders the local input, the remote track's spectrum, and the post-processed output simultaneously. DSP-accurate collision markers show exactly where the algorithm detects and resolves frequency masking.

Nominal-CPU-Tax Design Built for high track counts, the plug-in features a fully lock-free audio callback and offloads all heavy FFT analysis to a dedicated background thread, ensuring your DAW remains responsive.

Per-Pair Controls When linking tracks, these controls apply strictly to that specific relationship, ensuring localized control over the unmasking behavior:

• Depth (0–12): Dictates the aggressiveness of the gain reduction. Low values yield gentle presence management; high values produce deep, surgical separation. Determines the deepest dB cut your largest notch will target.
• Q (0.5–8.0): Scales the bandwidth of the dynamic cuts above the frequency-optimized baseline.
• Sensitivity (0.2–4.0): Modifies the detection threshold. Higher values widen the detection net to catch subtle overlaps, while lower values demand a tighter, more direct frequency collision to trigger processing.
• Attack & Release (0.1–4.0×): Multipliers that fine-tune the adaptive timing baselines, allowing you to tailor the notch response to the specific envelope of the source material.
• Bands: Restrict masking resolution to specific frequency zones (Lows, Low-Mids, High-Mids, Highs, or All).

Per-Track Controls:

• Bass (0.0–2.0): A multiplier on the dedicated Bass Slot. Lower value results in a deeper cut, meaning steeper Bass ducking (0.0 = full depth, 1.0 = 50% depth, 2.0 = no depth).

Technical Specifications

Platform: Windows and MacOS (VST3 / AU).

Framework: JUCE.

Analyzer: 80 log-scale bands, 20 Hz – 20 kHz.

Filter Bank: 8 slots per instance (7 dynamic notch + 1 dedicated Bass slot).

Max Instances: 64 per session.

Audio Thread: Fully lock-free, zero heap allocation in the audio callback.