Audio effect plugin

Software that processes audio inside a host DAW — the mixing/production counterpart to a hardware synthesizer. Rather than generating a signal, it shapes one: delay, filtering, saturation, dynamics, and so on. The founding corpus has two, at opposite ends of the design spectrum.

Plugin formats

The interoperability layer is a small set of formats a host loads: VST3 (Steinberg), AU (Apple desktop), AUv3 (Apple’s iOS/iPad sandboxed format), and CLAP (the newer open format). shinronin ships VST3/AU/CLAP on desktop plus AUv3 on iPad; next-oxx is VST3-only and Windows-only. Format and platform coverage is itself a signal of how much a free plugin is meant to travel.

A different kind of format is Max for Live (M4L) — a device built in Max that runs inside Ableton Live rather than as a standalone cross-host plugin, so it needs Live plus an M4L license. beast-repeat is the spoke’s M4L example: host-locked in exchange for deep integration (it rebuilds and extends a stock Ableton device), where a VST3 is host-portable but standalone.

M4L is one of two ways a patch reaches a DAW, and the other has the opposite trade-off. plugdata (free) wraps Pure Data patches as plugins any host can load, so a Pd patch travels like a VST3 while an M4L device stays in Live — same authoring idea, one host-locked and one not. Remo De Vico distributes granular Pd patches this way. What the wrapper route gives up is the polish of a compiled plugin; what it buys is that the maker never leaves the patching canvas.

Two design poles

• Open and modular — shinronin exposes a patch matrix over its delays/filters/ saturators, so the user builds the effect. Reborn from a 2003 plugin, now freeware.
• Fixed and specific — next-oxx does one thing: OTT-style multiband compression, upward and downward compression across frequency bands at once, for the aggressive dubstep/DnB sound. No routing to configure; the value is a tuned, artifact-controlled engine.

These effects — shinronin, next-oxx, and beast-repeat — are free, the recurring pattern in this corner: software given away to build a maker’s audience, whether a studio reviving old IP (audio-damage), a label seeding producers (Next OXX), or an individual M4L author (dillon-bastan). Distinct from speech-audio-wiki, which owns neural audio models; a DSP effect lives here, an AI audio model there.

The batch-release / AI-generation wrinkle

Xnullx (2026-07) stretches the free-plugin pattern in a new direction: five free plugins shipped at once by a brand-new developer — a multiband granular processor (Grainbrain), multiband saturation (O2), a beat-slicer (Bunka), a wavefolder (Foldspace), and a granular freeze/smear (Halide), all VST3+AU only. The reviewer suspects they’re AI-generated, reading the generic site and interfaces and the all-at-once drop as the tell, with reader reports of install bugs and a CPU-hammering site. So “free” now has a low-effort/high-volume shadow: where the earlier freeware was a deliberate audience play by a known maker, this is plausibly machine-generated shovelware of uncertain quality — a failure mode worth watching as generation tooling lowers the cost of shipping a plugin to near zero.