How I Create a MIDI File
This is not a tutorial on MIDI sequencing; there are many
good basic ones on the web, but it does deal with how I personally
approach creating a MIDI file. I am using the first movement of the
Elgar Violin Concerto as an example.
Since I've done several Elgar files, I have now chosen the
Violin Concerto. I've decided to stick to 16 channels as Elgar has
reduced his orchestra so as not to drown out the solo violin, and also
so that this may be used as a file for a student to play along with
after muting the solo track ("a la Music Minus One"). But, I suspect it
would be hard to play along with because of the many tempo changes.
To save work, I choose my setup from another file I've
already finished, in this case, the 16 channel version of the Enigma
Variations. I "save as" "eevlnc1.mid" (and thereby keep the original
Enigma file intact). I then proceed to delete all the measures I've
created except the first two measures which contain the system
exclusive messages and the setup for each track. I also delete the
tempo track and meter tracks and change the tracks which differ from
the "Enigma" file.
I would be cautious if you are using a graphic sequencer
such as Encore, Finale and I think Sibelius since excessive cutting,
copying and pasting can corrupt the file. I'm using Studio Vision Pro
which is non-graphic but I still save (or should save) often,
just in case. In addition, I keep two earlier versions and the most
recent one in my online storage or a CD-RW.
I set my tempo at the composer's metronome mark to begin
with but sometimes do the "song header" faster or slower so as not to
drag out the beginning measures too much or, if it's too fast, the
system messages may not be received. A Romantic composer like Elgar (or
even Beethoven for that matter!) will not expect one tempo to be
slavishly followed. In this case, I will not keep a steady tempo but
change each more or less slightly according to various criteria such as
musical character or to point up certain notes.
Since this page has been available for some time and since
I've finished sequencing the Concerto, I've decided to concentrate now
on how I've done the "Cadenza accompagnata" which comes close to the
end of the last movement and is the most difficult to "midify". It also
sums up most of the problems that one encounters elsewhere in the work.
At #101, the beginning of the cadenza, the string sections
are divided into two each and this lasts through the bulk of the
cadenza. In this file and in the combined version (now posted.), I've
added these tracks as "divisis" below each section's regular tracks. At
first, the divisions are of the string ensemble which are marked
"ponticello" (at the bridge.) I find this effect is best achieved with
the controller 71. Putting it around 80 (default is 64 of course and it
should be returned to this value after the passage in question.) gives
that glassy quality and, despite opinions to the contrary, you are
risking some "MIDI clog" if you don't use the tremolo patch (i.e. write
out the tremolos.) where it's used in multiple channels simultaneously.
The following violin solo has a couple of measures marked "accel."
which seems to be ignored by almost all performers including Menuhin
and the composer but I see no reason why it shouldn't be observed as it
is effective with the ritard following.
In a cadenza such as this, there are many stretches which
are unmeasured such as the section at #102. As these are usually put
into measures in a MIDI file, these might be done as logically as
possible bearing in mind that 5/4, 7/4 and other uneven time signatures
are quite possible. Long notes and fermatas are best written out at
convenient note values so the pizzicato tremolos don't get too slow. As
long as the tempo changes are not too extreme, these changes should not
be audible. The "pizz. tremolandos" are done by plucking the strings
with 3 or 4 fingers and here a MIDI clog is also a danger. But, by not
doing the tremolos too fast and offsetting the notes in alternate
tracks, there is some improvement while keeping the "aeolian harp"
But I've noted for a while that there are a couple of spots
in the cadenza that didn't play right in my other software but I
finally fixed up the file and eliminated some MIDI clog by changing the
violin 1 divisi track to the tremolo patch instead of fast pizzicato
patches in that track, and changing the sound of it as well as I could
by using various controllers. It is hoped that it won't sound too bad
if the software or sound source doesn't respond to these controllers.
(As of 11/26/03 I have also posted an mp3 version which sounds more or
less properly.) Again, even after all of this, the file still doesn't
play right on a friend's (Dell) computer and I'll have to (reluctantly)
change the other tracks as I did the violin 1 divisi track.
I also want to point out that each performance "lines up"
the tremolo chord changes differently and I've done these fairly close
to the composer's performance.
There are a few measures in 4/4 time which have notes that
are marked with a fermata (corona.) and I've generally changed these to
5/4 to draw out the held notes. The actual duration needs to be changed
according to how much space is desired before going on. If the solo is
accompanied by non-tremolo chords or unaccompanied, then even greater
freedom of tempo is possible.
The single passage in harmonics is not a problem as long as
the notes are notated in the 7th octave (standard.) which automatically
gives that sound. If, in other works, the harmonics don't occur in that
octave, using pitchbend and other controllers may work quite well. In
the slow movement of the Elgar First Symphony (q.v.), I've taken
various approaches to this problem but mostly use pitchbend in one form
Trills are a special problem in many places. When they are
desired to start slower, this may be controlled by tempo. Often, the
first note in the changes of a lengthy trill passage should be
lengthened to show the change.
I have decided to add the following to this page to reflect
my current project which is a revision of my preexistent MIDI file of
the slow movement of the Brahms First Symphony.
In looking at the original file, I am struck by how much I
tried to get away with in terms, mostly, of expression events. Yes,
I've found various wrong octaves in a few spots which I corrected and
quite a few wrong notes and omissions. But basically the overall effect
to me was a lack of sonority and the general volume level which failed
to match that of the new first movement file. The third movement,
except for a few balance problems is generally more satisfactory and
hasn't require that same amount of revision.
After several years of creating these files, I've been
finding it more necessary in lyrical sections to shape the main melodic
lines as well as the secondary lines with the expression controller and
even the tempo track in order to highlight important notes by holding
them more or less longer. Working with the solo track of the Elgar
Violin Concerto taught me a lot here as I observed that I could get
much "change of color" on notes using these means as well as modulation
At the very beginning of this slow-movement file, I've
avoided a definite attack on the first chord by moving it backward and
putting in a crescendo from silence with CC11.
In the violin solo of this file, I did use the mod wheel but
have altered it by using the random tool to give more of a vibrato
effect. (In my software, I haven't found using pitchbend instead of the
mod wheel to be a good idea but please be aware of the possibility.)
However it seems impractical to use the mod wheel controller much in
the many orchestral patches so I find system messages at the beginning
of these tracks generally have to suffice.
Next page: about slides and