Dave Phillips wrote:
Greetings:
Ah, one of my favorite threads. I compose a lot with MIDI, and
sometimes I like to try making the results sound at least a little
more human. A few (probably obvious) notes :
No human player can play two notes in a row with exactly the same
precision, it's simply not possible. Small-range random changes to
velocity at least help shake up the flatness of MIDI parts. However,
if the range of variation is too wide the randomizations sound pointless.
No human players play with metronomic accuracy, nor is such a thing
desirable from human players. However, the range of variation is a
significant consideration. Set drummers often play the bass drum,
snare, hi-hat, and cymbal in slightly different tempo relations,
creating the various grooves and feels that underlie musical styles.
A "groover" would be a cool piece of software, i.e., something that
shapes the existing music into another rhythmic style, but I don't
think it would be easy to get right.
IMO the best way to humanize MIDI parts is to hand-code their
velocities, note by note where necessary, which means the coder must
have a practical knowledge of performance factors. I also routinely
add a tempo track that loops an asymmetric group of fluctuating tempo
values, like a very narrow-range LFO applied to tempo. Again, if the
width of the range is too great the looseness becomes sloppy,
destroying the intended effect. Usually I keep a range of +/-4 clicks,
e.g. 120-121-122-121-120-119. If these values are applied to tempo
events at the level of 16th-note triplets they'll have a nice
"upsetting" effect on the rigidity of the sequencer's tempo.
Dynamics and tempo relativities are profound aspects of any engaging
performance, whether a player plans them in advance or they occur as
an improvised response to the music (and possibly to the other players).
Just some thoughts for further consideration.
Best,
dp
Dave, have you seen this software I've just found?
http://www.midi-plugins.de/mplug/mplug-hum.html
This is quite close to what I want to achieve.
After I've read your comments, and those by Frank and Stephen, I believe
that the possibilty to choose different probability curves (linear,
gaussian, exponential, reverse exponential) is the next feature to add.
--
www.cesaremarilungo.com