As configured, the Seed listens for audio, feeds it to a Fourier<T, N> instance (which handles the buffering, windowing, forward FFT, frequency-domain processing, inverse FFT), and then outputs the processed audio. The example frequency-domain process is defined in the denoise method. Try modifying it!
I've found the program works well when N = 4096 and laps = 4 (higher frequency resolution, greater latency), or when N = 2048 and laps = 8 (higher time resolution, less latency).
To compile, make sure you modify the Makefile so that the definitions of DAISYSP_DIR and LIBDAISY_DIR point to your installations of DaisySP and libDaisy. Then, as usual,
make clean && make && make program-dfu