Cabinet IR loader
Nave
Where the sound resonates — impulse-response cabinet simulation for guitar and bass DI.
AU · VST3 · Standalone · macOS & Windows · AGPL-3.0
About
A cabinet impulse-response loader for reamping guitar and bass DI tracks: load a cab (or full-rig) IR and Nave convolves it with your signal through a zero-latency partitioned convolution engine, then shapes the result with a simulated mic-distance control, post-convolution filters, dry/wet mix, and output trim. With no IR loaded it runs a unit impulse — a true passthrough out of the box.
Why the name?
The nave is the vast central body of a church — the space whose stone and air turn a voice into a sound. Nave does the same for your guitar: the cabinet is the body your amplified signal speaks through, and this plugin is that body.
Under the hood
A morph, decomposed. Each IR is split into a minimum-phase magnitude spectrum (cepstral transform) and a bulk delay (sub-sample cross-correlation). Blend interpolates log-magnitudes — a geometric mean, so shared resonances hold their level — and glides the delay like a mic in motion. A worker thread resynthesises; the audio thread never waits and never locks.
A convolver built for exchange. Stock convolution engines hard-reset when an IR changes. Nave's custom uniformly-partitioned overlap-save convolver keeps one shared frequency-domain input history and two filter sets, crossfading their outputs amplitude-complementarily — an identical republish nulls below −100 dBFS, and it nulls against direct FIR convolution below 1e-5 at every partition edge case.
Sessions that carry their sound. IR audio is embedded in the plugin state — gzip'd, raw, capped at 10 s per slot. Delete the source files and the session still renders below −80 dBFS of the original. Paths are kept as metadata; the sound no longer depends on them.
Alignment you can measure. FFT cross-correlation with parabolic sub-sample refinement recovers offsets of 7, 37.25 and 0.5 samples to within 0.1 samples, and auto-corrects inverted polarity — taking a 50/50 blend from −5.8 dB of cancellation to 0.0 dB.
Honest engineering, enforced by CI. 120 tests on macOS and Windows: zero audio-thread allocations with every feature engaged (the allocator is replaced to prove it), zero reported latency in every configuration, per-sample gain ramps, measured filter slopes (−24.40 dB/oct where it says 24), and golden-render nulls that pin every pre-upgrade session — blend engaged or not — to bit-identical output.
Features
- Two independent IR slots — load any WAV/AIFF impulse response; loading happens off the audio thread and never blocks playback
- IR Blend — move between IR A and IR B by crossfade, or by Morph, which rebuilds a single impulse response from the two so there is nothing left to comb-filter (v0.3.0)
- Zero-latency convolution — uniformly partitioned engine, no plugin delay compensation overhead
- Distance — simulated mic-to-cab distance (proximity-effect bass and air-absorption darkening)
- LoCut / HiCut — post-convolution filters (20–800 Hz / 2–20 kHz) with explicit off positions
- Mix & Level — dry/wet (default fully wet) and ±24 dB output trim
- IR Align — Precise cross-correlates the two IRs to a fraction of a sample and flips a polarity-inverted capture automatically; Legacy keeps the pre-v0.3.0 onset detection
- IR Gain Match — level loaded IRs by raw energy or by K-weighted loudness, so swapping IRs compares tone rather than volume (v0.3.0)
- Min-Phase per slot — non-destructive minimum-phase transform so IRs from different sources blend without phase cancellation (v0.3.0)
- IR B Trim / Polarity / Delay — balance, flip and time-offset the B branch alone, ±5 ms (v0.3.0)
- Distance Air — adds mic-distance time of flight to the wet path, gliding rather than stepping under automation (v0.3.0)
- LoCut / HiCut slopes — 12 or 24 dB/oct, crossfaded over 10 ms so switching is silent mid-take (v0.3.0)
- Factory presets & preset management — browse/save/import/export via the in-plugin preset bar
- State recall embeds the IR audio itself — up to 10 s per slot — so a project reopens with the same cabinets even if the source files moved or are on another machine (v0.3.0)
Download
The latest builds are published on GitHub Releases.
The macOS binaries are Developer-ID-signed, notarised by Apple and stapled — they install and open without Gatekeeper warnings. Windows builds are not yet Authenticode-signed; SmartScreen may warn about an unknown publisher.
Compatible DAWs
Any modern DAW that supports VST3 or Audio Units. Pro Tools (AAX) is currently not supported.
Screenshots
Coming with the next release.
The custom interface is currently in design — screenshots will appear here as soon as it ships.
Audio examples
Coming with the next release.
Before/after clips are being recorded — audio examples will appear here with an upcoming release.
Support development
Basilica Audio is free software, built at night and tuned by ear. If it earns a place in your session, you can help keep the candles lit.
Donation links are not wired up yet — they will go live soon.






