This paper describes the implementation of a digital hardware architecture for correcting the effect produced by group delay distortion on multiway loudspeakers. The correction is performed by a digital filter implemented in the form of an FFT convolution. The application imposes real-time execution on an embedded low-cost platform using a high-resolution audio format (stereo coded on 24 bits @ 96 kHz). The result is a pipelined streaming architecture composed of FFT-complex multiplication-iFFT performed on 32,768 samples using a floating-point representation. The filter computation is performed in 72 ms and the overall maximal audio latency is 169 ms running on a Cyclone V FPGA.