Daniel Iglesia

I've had some people asking for the code for a crude audio morph I made

recently, so I figure I'd post it far and wide:

http://www.music.columbia.edu/~daniglesia/software/morphPFFT.zip

(includes windows (cygwin) compiled class files, help patch, and pfft~

It's java code for mxj~, made to run within a pfft~ subpatcher.

It takes two audio streams, a percentage value (0. to 1.), and a threshold

value. For each pair of input frames, it looks at every bin with magnitude

above the threshold value, creates a series of correlations between these

non-trivial bins, and creates a new FFT frame with a percentage

interpolation between the two. A low threshold value means more non-trivial

bins, which means that the interpolations have less distance to travel and

therefore a more subtle result; inversely, a higher threshold value means

fewer nontrivial bins and greater aural distances to interpolate over,

yielding a more active result. It's crude in several ways, primarily because

it is interpolating bins, not frequencies (which are represented by several

neighboring bins); so you won't get smooth glissandi on spectrally simple

audio, but a more chaotic hodgepodge of nearby frequencies. You can also

extrapolate beyond percentage values of 0. and 1. for cool noisy effects.

I've had some people asking for the code for a crude audio morph I made
recently, so I figure I'd post it far and wide:

http://www.music.columbia.edu/~daniglesia/software/morphPFFT.zip 

(includes windows (cygwin) compiled class files, help patch, and pfft~
subpatch example)

It's java code for mxj~, made to run within a pfft~ subpatcher. 

It takes two audio streams, a percentage value (0. to 1.), and a threshold
value. For each pair of input frames, it looks at every bin with magnitude
above the threshold value, creates a series of correlations between these
non-trivial bins, and creates a new FFT frame with a percentage
interpolation between the two. A low threshold value means more non-trivial
bins, which means that the interpolations have less distance to travel and
therefore a more subtle result; inversely, a higher threshold value means
fewer nontrivial bins and greater aural distances to interpolate over,
yielding a more active result. It's crude in several ways, primarily because
it is interpolating bins, not frequencies (which are represented by several
neighboring bins); so you won't get smooth glissandi on spectrally simple
audio, but a more chaotic hodgepodge of nearby frequencies. You can also
extrapolate beyond percentage values of 0. and 1. for cool noisy effects.

audio morph in mxj~