I have been quietly building an audio plugin suite for the past year. It is not a synth, not a sampler — it is a different way of thinking about filters entirely. The first plugin is an 8-band equalizer, and it already does things no commercial EQ I have tested can replicate.

Four Ways to Hear a Band

Most EQs operate in one domain: amplitude. You boost 2kHz by 3dB, the gain multiplier scales the waveform, and you move on. But amplitude is only one way to slice a signal.

The plugin I built operates in four distinct processing modes. I am not going to explain the internal representation — the point is how they sound:

Surface — the baseline path. This is what you expect from a conventional digital EQ: precise, phase-aware, predictable. I use it when I know exactly what I want and I do not want surprises.

Resonance — the filter sees the signal's sustained presence rather than its instantaneous value. Harmonic content behaves differently here. A 3dB boost in this mode feels heavier than amplitude space — the spectrum thickens in a way that does not read as mere loudness. It is the mode I reach for when a track sounds thin but boosting does not help.

Depth — a wide-angle view that changes how the filter interprets dynamic range. This is the cleanest mode for broad tonal shaping. You can cut 200Hz aggressively and the result stays stable even on dense program material that would break a conventional minimum-phase filter.

Motion — the band follows the signal rather than sitting still. Loud passages get one treatment, quiet passages get another, all from a single parameter set. The tracking is fast enough to hold onto transients without pumping. I use it on drums and acoustic instruments where a static EQ either chokes the attack or leaves the body unstitched.

The Pre-Ringing Problem

Linear phase EQ sounds clean — until you hit a sharp transition. A steep highpass or a narrow notch creates pre-ringing: a faint ghost of the signal appears before the transient. This is not a bug in the filter implementation. It is the Gibbs phenomenon, baked into the mathematics of any linear time-invariant system with a brickwall-like transition.

I spent months on this. Conventional wisdom says you trade between pre-ringing (linear phase) and phase distortion (minimum phase). I wanted neither.

The answer turned out to be a change of listening frame. Not a different filter topology — a different way of asking the filter what it is hearing. By operating in a mode where the sustained presence of the signal is the native quantity rather than an afterthought, the filter response behaves differently. The transition curves smooth out in a way that does not require a linear-phase impulse response, which means no time-domain symmetry, which means no pre-ringing tail.

I am not going to publish the derivation. The point is not the formula. The point is that there exists a way of posing the filtering problem that makes IIR filters behave like they have been given an extra degree of freedom.

The Reverb Plugin

Once the EQ framework was solid, I started on a reverb. I wanted a feedback delay network that could do two things most algorithmic reverbs cannot:

First, a damping law that adapts to level. In a real room, the high-frequency decay rate depends on how loud the reflection is — loud early reflections hit air and surface absorption harder than quiet late tail. Most digital reverbs use a fixed damping curve. I built a mode where the damping is implicit in the signal domain itself: loud passages automatically get stronger HF roll-off, quiet passages stay brighter. No sidechain, no envelope detector — the property falls out of the math.

Second, no explicit clipper in the feedback path. Most FDN reverbs need a tanh or limiter inside the loop to prevent exponential blow-up. That nonlinearity colors the tail and thins the decay. My architecture avoids this by keeping the feedback bounded through structure rather than saturation.

What Comes Next

The EQ is in closed beta on my own mixes. The reverb is functional and passing its test suite. The next plugin on the roadmap is spatial positioning — not a panner, but a room simulation that treats position as a continuous field rather than a left-right fader.

I want a plugin where moving a source "closer" changes not just level and pan law, but the relative weighting of transient energy versus diffuse energy, the apparent size of the source, and the HF damping of the direct path — all from a single XY control. This is what analog consoles and high-end outboard gear do intuitively. I am trying to make it mathematically rigorous without losing the feel.

All of this is built in Rust. The DSP is zero-latency, sample-by-sample, no FFT anywhere in the chain. The plugins are VST3 and CLAP. The GUI is custom wgpu. I am not releasing anything until it survives a full mixing cycle on a real track.