18 Mar
2003
18 Mar
'03
1:18 p.m.
Hi,
Have any of these soft synths that use boxes and user-based connectivity
(SSM, Octavian, gAlan, Beast, etc.) addressed the issues of polyphony yet?
Mark
-----Original Message-----
From: linux-audio-user-admin(a)music.columbia.edu
[mailto:linux-audio-user-admin@music.columbia.edu]On Behalf Of Mike
Rawes
Sent: Tuesday, March 18, 2003 8:11 AM
To: linux-audio-user(a)music.columbia.edu
Cc: alsamodular-devel(a)lists.sourceforge.net
Subject: Re: [linux-audio-user] Modular synths of the world, unite and
take over :-)
--- Lukas Degener <AFBLukas(a)gmx.de> wrote: > Roman Kaljakin wrote:
Sseems we have a little problem.
Ok, maybe problem is not exactly the right term, let's put it this way:
Right now, there is beast, there is gAlan, there is ams, and of course
there is also pd, jmax, etc. and lately Octavian, and, rats,
there maybe
a zillion of other apps out there that i do not
yet know of (authors of
such apps, please append your project to my list:-) )
There's SpiralSynthModular (SSM), which I use most frequently
(I've not tried
Beast/BSE, nor Octavian).
While each of this projects may have unique
approaches to certain
problems, or use different metaphors for similar things, i
guess it is a
valid assumption, that they all have _very_
similar goals in terms of
functionality. At least that was my impression after talking
with Stefan
(beast) and Torben (gAlan).
The Big Question(tm):
How can we avoid redundant work?
My (somewhat utopious) suggestion:
maybe we should think in components that use/modify a common
datastructure/model.
Here's my take on this...
----
1. A unified Plugin API
All of the above softsynths have their own plugin format, for
various reasons.
LADSPA (http://www.ladspa.org) has helped with this, making much
excellent code
available for all these softsynths. However, the (intentional)
simplicity of
LADSPA has prevented the various formats being replaced entirely.
The Generalized Music Plugin API discussions
(http://www.freelists.org/archives/gmpi) are working towards a more
encompassing API, but it looks to be a long way off.
----
2. A single codebase for building, representing and running graphs (AKA
networks, patches) of plugins.
Almost all of these softsynths are pretty much exactly the same 'under the
hood' - just a connected bunch of plugins executed in a
particular order. The
only real obstacle to unifying this is the plugin API issue above.
I'd certainly like to see a unified engine for LADSPA - I've been
on-and-off
planning one, but SSM is so close to what I'd like that I'm not motivated
enough :/
My thoughts on engine-things so far (this is all a bit of a brain dump,
apologies:)
No GUI - all functionality exposed through various forms of
Inter-Process Communication (IPC):
Graph State Query and Manipulation (command set for adding
plugins, making connections etc).
Audio I/O - JACK (http://jackit.sf.net) would be the choice here
Control - something like the LADSPA Control
Protocol (http://www.op.net/~pbd/lcp.html)
would be good.
Built-in support for subpatches:
The graph itself could be kept 'flat', with some sort of
abstract representation of subpatches on top of it. Including
ability to 'register' subpatches as plugins.
The thought of being able to set up, for example, a mixer
channel out of multiple LADSPA plugins, register it as a plugin,
and then add sixteen of them to a graph to make a mixer has a
certain appeal :)
----
3. Separation of graph/engine and any UI.
Any UI would communicate with the engine using IPC - this would allow
multiple views of the network - e.g. a wiring diagram for setting up,
a set of knobs'n'sliders (real or virtual) arranged nicely for
performance, commandline scripting, or any combination...
-
Mike
This has almost (but not completly) nothing to do
with merging ;-)
It means that we have to agree on what exactly that common model would
be, and after that, we would go on writing
different components that actualy work with this model. The problem of
course would be to define that model.
But once we have managed this, it should be possible to have different
components cooperate on the same instance of that model at the same
time, without knowing about each other, i.e. all "inter-component"
communication would be established via the model. I tend to think of
this as a "Macro Model/View/Controller" pattern. (i have forgotten the
actual term, something like "Document Oriented Design" i think.)
So this goes out to Matthias, Torben, Stefan, Roman, and of course
everyone else interested in such a thing:
What do you think?
Regards,
Lukas
__________________________________________________
Do You Yahoo!?
Everything you'll ever need on one web page
from News and Sport to Email and Music Charts
http://uk.my.yahoo.com