Post by JohnG on Mar 7, 2012 15:46:53 GMT
Just thought I'd drop a few notes on how I went about creating the Handel's Largo tracks that I posted recently.
Unlike many musicians here, I believe, I work primarily from notation, more frequently referred to as 'the dots' by many I think.
I suppose that's because, when I was learning to play, I was taught in the classical manner, reading from a score.
So, in order to fulfill the request, I visited first a web site that hosts pdf's of music that is now out of copyright. Its name is IMSLP. It's here: imslp.org (yep no http or www).
If you go to the Petrucci Music Libary (just add a "/wiki" to the end of the "org" and you can search for all sorts of things, not least composers' names and the title of a piece of music by them.
So "handel largo" turned up a list (done with google search engine) the first entry of which was to Serse (or Xerxes) the opera which the aria comes from, and that takes us here: imslp.org/wiki/Serse,_HWV_40_(Handel,_George_Frideric).
There are many copies here and, after downloading and looking at quite a few, I carefully selected one of the full scores of "Ombra mai fu" (the name of the largo), and the complete set of individual parts (although I only need the 2 violins, viola and cello/basses/continuo). Then I searched for a realisation for the 'figured bass'. I found several amongst the 'Arrangements and Transcriptions'.
For those who've never come across a figured bass before, it used to be a way that the composer could instruct the organ or harpsichord player what chords etc. should be extemporised around when playing the accompaniment. A good player of that time could sight read the numbers under a bass note (the figures) and make up an accompaniment to suit as he or she played. More or less a lost art.
Having acquired a set of scores to work from, I then run up Mozart Notation software and enter the notes for one instrument. (I use Mozart because I find it incredibly quick and easy to enter the notes using it. A lot faster than anything else I've tried.) I then export this as a MIDI file and import it into XGworks sequencing software. Just the one track, initially.
I do various bits of sequence set up with the first track and then go back to Mozart and gradually add each instrument track, one at a time. Score, MIDI. Why don't I do all instruments at the same time? Well, I used to. But often I'd get bogged down when trying to complete a big orchestral piece, and I found this way suited my temperament better. Score, MIDI. Score, MIDI. Just me.
Having got all the notes into MIDI format I then set an overall tempo at the start of the piece, and the channel volume of each channel to 100, and listen, first a track at a time and then gradually more and more together. This is basically to hunt out any typo mistakes I might have made.
But this last statement begs the question "listen how?" If the piece is 16 channels or less, then I run up a stand alone copy of the Virtual Instrument Sample Player ARIA, and load into each of the 16 slots a copy of the instrument I need. violin or flute or what have you. The instrument samples come from the Garritan Personal Orchestra. One of the best moderate cost orchestral sample libraries, in my opinion.
If more than 16 channels, then I run up a copy of VST Host and load the number of instances (usually just two) of ARIA that I need. Then add the instruments.
I leave the other controls, apart from reverb send (CC#91) alone. ARIA sets its own pan etc. when the instruments are loaded so, unless I want to change their default orchestral position, I leave these as set. Reverb send I set initially to 40 (out of 127), i.e. not too much reverb. I also choose the 'Room type' in ARIA's reverb settings.
XGworks sends the MIDI, via MIDI Yoke, to ARIA when I hit play. The sound goes to my E-MU 1616m card and thence to speakers and a small sub-woofer.
If all sounds in tune, then the next phase is to look at the score and implement the basic dynamics. With the first instrument, in Staff View, I highlight the notes of the first part to be played quietly (marked "p" in the score) and using a "velocity modify" macro in XGworks (called a 'job') I make the notes 'piano', that is I give them a velocity of 40 (initially at any rate). Then the next bit is forte, (marked "f") so I highlight those notes and they get set to 88.
Here's the table I work from (taken from a Yamaha document):
ppp=1-15 centre 8,
pp=16-31 centre 24,
p=32-47 centre 40,
mp=48-63 centre 56,
mf=64-79 centre 72,
f=80-95 centre 88,
ff=96-111 centre 104,
fff=112-127 centre 120.
I use centre values to start with.
And so I progress through the piece instrument by instrument getting 'note on' dynamics moderately correct. Often it doesn't need a lot more tweaking.
The Garritan Orchestra is different to a conventional MIDI instrument in that note on velocity does not dictate note volume for sustaining instruments i.e. woodwind, brass and strings and the organ. Just the amount of attack, the amount of spit (for brass) or tongue (for woodwind) or bow (for strings) etc.
I have to go again through each instrument part adding an initial Expression controller (CC#11) in front of each change of dynamic to get the volume (and timbre as it happens) at the correct level. I use the same table as above.
Then where there are hairpins (crescendos and diminuendos to use the 'correct' terminology) one can either draw into the controller pane a CC#11 curve, or better use a controller fader on a MIDI keyboard controller. The latter sounds more realistic.
Then I look at phrasing. Some notes will have a curve over a group of them from time to time in the score. These notes should have just one 'attack' sound at the start and the rest should blend more smoothly together. Just like a wind player would play a group of notes with one continuous breath. This is known as playing legato. So the length of each note needs to be extended to overlap the start of the next note in the phrase. This is usually easily done in the Piano Roll View (the PRV), where the right hand end of a note can be dragged a little further right. Then a 'play legato' MIDI controller event needs to be entered as appropriate.
And so one adjusts the way that a musician plays, legato, staccato, pizzicato and so forth, instrument by instrument by adjusting note length, or the timing between note on and note off, and also adjusting the expression of the whole track by creating 'expression curves' (either drawn in or with a controller).
With that completed for every instrument one now has something that is beginning to sound more realistic. A closer examination of tempi. A comparison of one instrument playing with another may lead to some adjustments to expression or overall channel volume.
Then one might look at a part and decide that alternate notes should sound slightly different as e.g. the cellist draws his bow back and forth across the strings. Many of these sample libraries have two different samples recorded for each note, one for up bow the other for down bow. One might then add a 'key switch' to trigger the alternation between them, or another to switch between a vibrato and a non-vibrato sample.
A 'key switch' is a MIDI note that is inserted into the track which lies outside the acoustic range of the instrument being 'played'. It is often a very low note. No sound is triggered. It just tells the sample player to use a different set of samples.
To give just one example, all three of the solo violins have 12 articulations to choose from each selected using a key switch.
One may feel that some notes should be pulled forward fractionally in time, or played slightly late. At the end, each channel should sound moderately like a real musician playing. If one can achieve that then one has neared completion.
Final adjustment of relative volumes, adjustments to tempi and so on take place until one is happy with the result. As do reverb and other effects.
Have I missed anything? Probably.
I'll review it in a day or so and maybe add some more, or change bits and pieces.
Any use to anybody?
JohnG.
Unlike many musicians here, I believe, I work primarily from notation, more frequently referred to as 'the dots' by many I think.
I suppose that's because, when I was learning to play, I was taught in the classical manner, reading from a score.
So, in order to fulfill the request, I visited first a web site that hosts pdf's of music that is now out of copyright. Its name is IMSLP. It's here: imslp.org (yep no http or www).
If you go to the Petrucci Music Libary (just add a "/wiki" to the end of the "org" and you can search for all sorts of things, not least composers' names and the title of a piece of music by them.
So "handel largo" turned up a list (done with google search engine) the first entry of which was to Serse (or Xerxes) the opera which the aria comes from, and that takes us here: imslp.org/wiki/Serse,_HWV_40_(Handel,_George_Frideric).
There are many copies here and, after downloading and looking at quite a few, I carefully selected one of the full scores of "Ombra mai fu" (the name of the largo), and the complete set of individual parts (although I only need the 2 violins, viola and cello/basses/continuo). Then I searched for a realisation for the 'figured bass'. I found several amongst the 'Arrangements and Transcriptions'.
For those who've never come across a figured bass before, it used to be a way that the composer could instruct the organ or harpsichord player what chords etc. should be extemporised around when playing the accompaniment. A good player of that time could sight read the numbers under a bass note (the figures) and make up an accompaniment to suit as he or she played. More or less a lost art.
Having acquired a set of scores to work from, I then run up Mozart Notation software and enter the notes for one instrument. (I use Mozart because I find it incredibly quick and easy to enter the notes using it. A lot faster than anything else I've tried.) I then export this as a MIDI file and import it into XGworks sequencing software. Just the one track, initially.
I do various bits of sequence set up with the first track and then go back to Mozart and gradually add each instrument track, one at a time. Score, MIDI. Why don't I do all instruments at the same time? Well, I used to. But often I'd get bogged down when trying to complete a big orchestral piece, and I found this way suited my temperament better. Score, MIDI. Score, MIDI. Just me.
Having got all the notes into MIDI format I then set an overall tempo at the start of the piece, and the channel volume of each channel to 100, and listen, first a track at a time and then gradually more and more together. This is basically to hunt out any typo mistakes I might have made.
But this last statement begs the question "listen how?" If the piece is 16 channels or less, then I run up a stand alone copy of the Virtual Instrument Sample Player ARIA, and load into each of the 16 slots a copy of the instrument I need. violin or flute or what have you. The instrument samples come from the Garritan Personal Orchestra. One of the best moderate cost orchestral sample libraries, in my opinion.
If more than 16 channels, then I run up a copy of VST Host and load the number of instances (usually just two) of ARIA that I need. Then add the instruments.
I leave the other controls, apart from reverb send (CC#91) alone. ARIA sets its own pan etc. when the instruments are loaded so, unless I want to change their default orchestral position, I leave these as set. Reverb send I set initially to 40 (out of 127), i.e. not too much reverb. I also choose the 'Room type' in ARIA's reverb settings.
XGworks sends the MIDI, via MIDI Yoke, to ARIA when I hit play. The sound goes to my E-MU 1616m card and thence to speakers and a small sub-woofer.
If all sounds in tune, then the next phase is to look at the score and implement the basic dynamics. With the first instrument, in Staff View, I highlight the notes of the first part to be played quietly (marked "p" in the score) and using a "velocity modify" macro in XGworks (called a 'job') I make the notes 'piano', that is I give them a velocity of 40 (initially at any rate). Then the next bit is forte, (marked "f") so I highlight those notes and they get set to 88.
Here's the table I work from (taken from a Yamaha document):
ppp=1-15 centre 8,
pp=16-31 centre 24,
p=32-47 centre 40,
mp=48-63 centre 56,
mf=64-79 centre 72,
f=80-95 centre 88,
ff=96-111 centre 104,
fff=112-127 centre 120.
I use centre values to start with.
And so I progress through the piece instrument by instrument getting 'note on' dynamics moderately correct. Often it doesn't need a lot more tweaking.
The Garritan Orchestra is different to a conventional MIDI instrument in that note on velocity does not dictate note volume for sustaining instruments i.e. woodwind, brass and strings and the organ. Just the amount of attack, the amount of spit (for brass) or tongue (for woodwind) or bow (for strings) etc.
I have to go again through each instrument part adding an initial Expression controller (CC#11) in front of each change of dynamic to get the volume (and timbre as it happens) at the correct level. I use the same table as above.
Then where there are hairpins (crescendos and diminuendos to use the 'correct' terminology) one can either draw into the controller pane a CC#11 curve, or better use a controller fader on a MIDI keyboard controller. The latter sounds more realistic.
Then I look at phrasing. Some notes will have a curve over a group of them from time to time in the score. These notes should have just one 'attack' sound at the start and the rest should blend more smoothly together. Just like a wind player would play a group of notes with one continuous breath. This is known as playing legato. So the length of each note needs to be extended to overlap the start of the next note in the phrase. This is usually easily done in the Piano Roll View (the PRV), where the right hand end of a note can be dragged a little further right. Then a 'play legato' MIDI controller event needs to be entered as appropriate.
And so one adjusts the way that a musician plays, legato, staccato, pizzicato and so forth, instrument by instrument by adjusting note length, or the timing between note on and note off, and also adjusting the expression of the whole track by creating 'expression curves' (either drawn in or with a controller).
With that completed for every instrument one now has something that is beginning to sound more realistic. A closer examination of tempi. A comparison of one instrument playing with another may lead to some adjustments to expression or overall channel volume.
Then one might look at a part and decide that alternate notes should sound slightly different as e.g. the cellist draws his bow back and forth across the strings. Many of these sample libraries have two different samples recorded for each note, one for up bow the other for down bow. One might then add a 'key switch' to trigger the alternation between them, or another to switch between a vibrato and a non-vibrato sample.
A 'key switch' is a MIDI note that is inserted into the track which lies outside the acoustic range of the instrument being 'played'. It is often a very low note. No sound is triggered. It just tells the sample player to use a different set of samples.
To give just one example, all three of the solo violins have 12 articulations to choose from each selected using a key switch.
One may feel that some notes should be pulled forward fractionally in time, or played slightly late. At the end, each channel should sound moderately like a real musician playing. If one can achieve that then one has neared completion.
Final adjustment of relative volumes, adjustments to tempi and so on take place until one is happy with the result. As do reverb and other effects.
Have I missed anything? Probably.
I'll review it in a day or so and maybe add some more, or change bits and pieces.
Any use to anybody?
JohnG.
