If 2150 iterations of real-to-complex FFT of size 2048 takes less than 10 seconds, then we may be able to make a high quality real-time phase vocoder (just).
Chris@1: Chris@1:A phase-vocoder of course must use overlapped Chris@1: windowed FFT (although you can choose the size, within limits), IFFT, Chris@1: and cartesian-polar conversion to calculate the phase for the Chris@1: instantaneous frequency.
Chris@1: Chris@1:A reasonable estimate of CPU cost for the whole thing is somewhere Chris@1: around 10x the cost of simple non-overlapping short-time forward Chris@1: Fourier transforms across the signal.
Chris@1: Chris@1:2150 iterations corresponds to 100 seconds of audio non-overlapped at Chris@1: 44.1kHz, so if that takes less than 10 second, then in theory we might Chris@1: be OK.
Chris@2: Chris@1: Chris@1: Chris@1: Chris@1: Chris@2: Chris@2: Chris@2: Chris@1: Chris@1: