>Also note that the strength of the note-on often affects the output, e g
>on a piano sound, a weak note has less treble than a strong one.
>
>> My apologies if I am somewhat lacking coherence here, I am still trying
>> to get a decent grasp on the subject matter.
>
>I'm not exactly sure what would be considered a part of your project or
>not, but have you considered actually having FluidSynth as your sampler
>(i e calling it in real-time from Supercollider), and if you want to
>apply some extra envelopes, modify the FluidSynth code in relevant places?
>
>That would probably be the simplest method if you want to accurately
>play back soundfonts. If you just want to get something easy to play
>with, you can use either method 1 or 2,
and implement as much as you want.
>
>// David
Thanks for your response , it is very instructive.
Here is what I hear you saying:
1.Soundfont generally make full use of the synthesis model described in the spec.
I.e. if one tries to use the wave tables but fails to implement certain parts of the spec, the
sounds won't be accurately reproduced.
2.Also, if I just sample fluidsynth output and apply an envelope it will not be an authentic
reproduction of the sound (treble example for piano).
Do I understand you correctly?
Suppose we decided to go ahead with sampling fluidsynth output for the moment to get things started.
Would using 5 wave tables (i.e. one per envelope phase) per pitch likely be a noticable improvement
over using a single one per pitch?
Do you happen to know whether there tends to be a correlation between wave tables and pitches in
soundfonts?
E.g. for a given soundfont it might be one wave table for every x pitches, or one table per pitch or
several tables per pitch? Or is it simply more complex/less orderly than that and there is no rule of thumb
describing the correlation between the number of wave tables for a given instrument and its pitches?
The main reason we haven't seriously considered routing fluidsynth output into suppercollider
are concerns that switching out sounds during a live performance would be a bit of a hassle if
it either involved replugging connections in qjackctl and bringing up a new fluidsynth instance
or typing something into a command line to load a new soundfont.
Best regards,
Michael