NeuralRack is an advanced neural model loader designed for audio processing. It enables loading two separate model files that can be run in sequence, with independent control over the input and output for each model. This flexibility allows for complex processing chains tailored to specific audio production needs.

The plugin supports a variety of file formats, including .nam, .json, and .aidax, through integration with the NeuralAudio engine. It also features convolution with impulse response files using an FFT-based convolver, and resampling is managed by the Libzita-resampler. Additionally, NeuralRack offers a six-band equalizer for precise tone shaping, enhancing the overall audio output.

NeuralRack provides a mono input and stereo output, allowing separate impulse response files for each output channel to create rich spatial effects. A buffered mode is available, introducing minimal latency by offloading neural processing to a background thread, which helps reduce DSP load without compromising performance. The latency introduced is communicated to the host for proper synchronization.