The engine

Mamba-3, per-sample,
in pure C++17.

A multi-head MIMO state-space model running per-sample in the time domain — no STFT in the model path, no external inference runtime, no GPU dependency at runtime.

Architecture

Specification

Six stacked SSM blocks with RoPE rotary embedding, trapezoidal discretization, FiLM conditioning, and MIMO rank-4 heads.

ParameterValue
ArchitectureMamba-3 SSM, multi-head MIMO R=4
d_model / d_state128 / 32
Layers6 (stack-allocated SSMBlocks)
Parameters~1.66M trainable
ProcessingPer-sample, time-domain
WeightsFlat binary model.bin, mmap, version-checked (V2)
DependenciesC++17 stdlib only
Latency0 samples reported
Verification

The test suite is the datasheet.

ClaimEvidence
CLAP conformanceclap-validator in CI
Cross-platform buildGitHub Actions matrix — Linux · macOS · Windows
Numerical correctnessBit-exact Python ↔ C++ parity (CTest)
RT-safetyZero heap allocation on audio thread; deferred atomic state reset; lock-free SPSC queue
Bypass behaviorClick-free 50 ms equal-gain crossfade
Stability60 s @ 44.1 kHz long-run, no NaN/Inf
Roadmap

R=8 spectral engine

The next generation adds an STFT frontend (n_fft=2048), complex-pole SSM, CREPE f0 conditioning, and cross-track attention at ~4.26M parameters. Training pipeline validated on datacenter GPUs; the C++ engine rewrite is in progress.

Latency disclosure: the spectral engine introduces 1024-sample algorithmic latency (n_fft/2). Current shipping engine: 0 samples. Both figures will be published with reproducible measurements.

Benchmark methodology →