@SgtChipo Even if you reduce the block size you still need 64 samples for the codecs.... so total 64/44100x2=2.9ms. With a block size of 1 you could make a small delay, but you still need to wait to push it through the dac.
Someone more knowledgeable than me (many on this forum) might know how to shift the phase via fft though (.... the same thing).
@katjav 's site would be a good place to start..... http://www.katjaas.nl/home/home.html
I think you will be disappointed even if you achieve small delays in your patch. You will still have a minimum of 2.9ms from input to output on your computer, but maybe that is acceptable (as a delay in the processing) and as you actually need only the difference between 2 channels..... not overall....... then maybe..........but I am not sure that it is possible in Pd.....!?
Dedicated 1-bit dsp hardware and assembly language are the usual requirements....