CPU of models (FEEDBACK)

[xoxos]

please reply to this thread!

1.)
tell me how you find the cpu load of my physical models.

2.)
is the cpu about right for one part, do you have to mix down, or could they use twice as much and it wouldn't bother you?

3.)
regardless of cpu use, how many lfos and envelopes would you like to have? (given that this must be a finite number).

4.)
are there factors such as gui size or appearance that factor in your user experience?

please allow me to include your interests in future development and take a minute to post

[xoxos]

[Az're-ul]

1.)
tell me how you find the cpu load of my physical models.

Not too bad for me but then again I have a quadcore.
Boing2 typically takes between 5 & 10%
Phybes uses 2 to 5%
Piamo uses 4 to 7%

2.)
is the cpu about right for one part, do you have to mix down, or could they use twice as much and it wouldn't bother you?

I could use twice as much without any bother now but I used to have problems when I had an Athlon XP 3200

3.)
regardless of cpu use, how many lfos and envelopes would you like to have? (given that this must be a finite number).

Things are complicated enough now without adding more controls Sorry can't really comment on this one but I'd vote for more if it improved on the already high quality of the sounds - I'd just need lots of presets already created for me.

4.)
are there factors such as gui size or appearance that factor in your user experience?

I'd probably opt for slightly bigger GUIs with more contrast given the choice but they're pretty good as they are.

Slightly OT FR - I find I quite often stick to what I know when I program presets with your synths so a randomizer function per section would be make things a little more interesting from my point of view.

Steve

[ponce]

please reply to this thread!

1.)
tell me how you find the cpu load of my physical models.

Excellent. You seem to be very concerned about CPU and that's great for popularizing physical models.

2.)
is the cpu about right for one part, do you have to mix down, or could they use twice as much and it wouldn't bother you?

In fact I wouldn't be bothered by 3x the CPU load provided there is more stability in your models.

Often your models have (besides too much parameters left to the user a lack of stability which gives a diffuse impression of reject. An example of this is in Cylinder33 or Snare. I know there are advantages of not stabilizing these models (purer sound, less CPU). But I have tinnitus and I hate these sudden burst of noise pain.

On the other hand, the fact is that physical models seems to be unstable near the most interesting areas of parameter space :/.

3.)
regardless of cpu use, how many lfos and envelopes would you like to have? (given that this must be a finite number).

It's up to you.

4.)
are there factors such as gui size or appearance that factor in your user experience?

please allow me to include your interests in future development and take a minute to post

Indeed, the GUI is quite small but I usually find it good when sliders have a "slow" mode when pressing CTRL for example.

As a global POV : excellent work. Some of your plugins are really ground-breaking. I'm currently interested in physical models and I'm impressed by your workflow and discoveries.

[xoxos]

i'm surprised to hear of someone else using those vst! they were produced in my first few months of modeling with code. new drums are my next project.

ctrl-drag should work with every control in synthedit

thanks! i don't know about ground-breaking. there's so much work in emulation that doesn't enter the market. people aren't really aware..

there are dozens and dozens of authors on physmodeling horn flares.. and how many vst developers?

and that's in english.. last week i spent a day going through papers on trumpet modeling in spanish at SEA.

[ponce]

There is for sure a lot of research work on physical models but I think a lot of these research are made without real-time constraints in mind or desire to create methods understandable by busy developers.

For a sample, I've read a PhD on realistic drum synthesis and it was like the man was trying to use every method to complexify his model, like using hundreds of unitary waveguides arranged like tetrahedron. Then after that formula nightmare, the provided result sample was ridiculously deceptive.

Also, maybe everyone subconsciously think that physical models are generally not CPU-efficient and good-sounding enough for serious use. I find them to be extensively used in minimal techno though (Extrawelt...).

[xoxos]

until 2006, threads on physical modeling uniformly reached 10 pages, sometimes running in parallel. then, i made physical models and the entire world subconsciously lost interest

watch what happens when i decide to make something else listen to my song!

before pling3, there were no free string models with tuned dispersion filtering, let alone orthagonal coupling. read any "pling3 is amazing" threads lately?

i think most people have other things to think about nowadays than plugins.

i deal in principle, eg. i studied latin and have a rudimentary apprehension of the array of modern languages because of it

these are the ideal forms real instruments aspire to be for me, i see the benefit artistically, socially/memetically, and entelechically, so i'm motivated. these are also the same filters i perceive sampling (duplication/source-reference/master-slave).