The list piporg-l [serving the pipe organ fans] had a link to an IEEE
magazine article on Zenph.  It presents the recreation of a Jerry Lee
Lewis stage performance, in which the piano actually cannot be heard, by
recreating his hands on the somewhat visible piano.  This is delightful
and it is quite different from the secret scientific data reduction
which Zenph has sold their clients.

Having maintained an interest in what is now called WAV-to-MIDI,
including participating in a special group at MMD to discuss it,
I realized that the pipe organ people in general have not been much
involved or interested in this, and wrote the following info-diatribe
which Robbie will cheerfully call a perfect example of "pure Karl."
Here it is --

 - - -

This technology is part wave frequency and signature analysis, part
data reduction and part careful archaeology with a good ear and eye.
With piano music, there is one voice, however busy the harmonic
overlaps might be.

I have worked on this off and on since 1961 with six generations of
hardware and software, ever hoping that the next advance will give the
key.  It remains, however, the holy grail of acoustic deconstruction.
Many physicists have simply stated and demonstrated why an audio record
is a priori too muddled to extract note events, notwithstanding that a
thoughtful musician can work out most of it.

The art of it comes in when combing through by iteration to match the
decoded reconstruction to the crude audio original.  The mind is a
teachable computer and comparator, so feeding the original signal in
one ear and the reconstruction in the other, the differences, though
subtle, become oddly apparent and the reconstruction can be modified
until the difference to the original approaches zero.

Zenph does not tell you that they do this, nor how they do it at all,
nor will they.  If it were simply engineering, they would have hundreds
of CDs already, so it clearly is something deeper.  They are having a
great time with doing it and teasing us with smatterings of results.

But with the organ, it is quite often more difficult to automatically
separate chords coming in from a cluster of ranks and perhaps several
divisions.  It is mud and distortion even if you had a microphone on
each chest.  If we have an accurate note record in the form of sheet
music which has been meticulously followed, we are miles ahead.

It is usually the unusual, the unique, the improvisational work that
one wants to decode, transcribe, and reproduce, and this is
astonishingly difficult even for the sensitive professional musician.
A sharp video of the musicians hands and feet could help, of course.
Anything -- please!

In most cases, it is much quicker and musically satisfying to have
a sensitive musician listen carefully and replay passages by ear,
emulating the sound and style of the recording.  This hand playing can
be recorded as a note and registration record and reproduced at any time.

There is a surprisingly vast and unappreciated literature of recordings
for several major reproducing organ systems, and these sound the same
as when they were recorded, yes, as notes and registration records,
a century ago.

Ask James D. Crank, who is the center of the copying, reconstructing
and reproducing of these ancient and wonderful systems.  His efforts
are saving this vast literature, unraveling the secrets of what
registrations were intended, and reproducing them on a carefully
organized version of each of the major systems.  It is delightful,
astonishing and praiseworthy.

I would sooner listen to this literature and ask that more recordings
be archived with it than wait, as the piano hopefuls have done, for
fifty years and only had three CDs of transcriptions to show for it.

Karl A. Petersen
Boise, Idaho