i started using NI Generator 1.5 ... (the predecessor to Reaktor) in 1998 and was in love! I used Generator for 20 years but found it hard
to get past a certain point in making my ideas come to life.

I tried a program called CPS that seemed cool.. but i made no progress. it looked like Pd and Generator had a child in windows95.

I tried 3 times over 20 years to learn Pure Data, but it never stuck/clicked until the last year... I credit youtube tutorials this time.

I've seen MAX and i think ill stick with Pd thanks,

I've seen a flowchart program for process automation that works the same way... no idea the name.

not visual, but i tried Csound... nope, not gonna happpen!

how about your experiences!?

Csound is my favorite of the audio DSLs but it is not really visual, while there are some front ends which give it a visual aspect they tend to ruin what really makes Csound great. One of my side projects has been working on a few application to make Csound more immediate and less calculated, the write/compile paradigm it has is not exactly fun but its way of scoring is very powerful.

because I have a fondness for languages which learning them is primarily about learning to work around their failings and quirks and vanilla pd provides that better than most everything else.

Agreed, and there's humor in those failings too. Pd's expressive poverty sometimes reminds me of one college homework exercise where we were asked to prove that 2 < 3 using only the axioms for integers and the definition of less-than. And that same poverty forced me to learn DSP more deeply than I would have otherwise.

there was one in Linux 20 years ago too, i think called ARTS... tried hard to use that one, but it was very early-days for the program.

I loved what Csound sounded like... but i couldnt do music that way.

That is why you use pd to control Csound through the Csound external/MIDI or use pd for score generation, that way you get the best of both worlds. Something about Csound makes it really easy to get good sounds.

ARTS was largely a failure from what I saw, KDE did not really know what it was doing and projects like AMS and Spiral Synth Modular did it better.

I've used Pure Data, Max and VCV Rack in the classroom, and don't really love any of them.

I use SuperCollider for my music. I'm not sure why but that hits a sweet spot for me... it "thinks" the way I do...? I see complaints that the architecture is confusing but tbh it made sense to me pretty much right away. You can 'do' loop large banks of signal chains even more easily than Max's mc. wrappers, and the way that graph optimization is atomized into smaller definitions (instead of globally like Pd) supports glitch-free incremental building up of a texture. And, having a full suite of data structures (esp. hash tables!) plus proper local function scope makes it feasible to implement more rich compositional algorithms with less fuss than in Pd or Max (yeah, I'd rather not implement my own call stack mechanism just to do real recursion, no thanks).

But when I tried to use SC in the classroom with mainland Chinese electronic music students, it was too many English keywords and they gave up. Using Pd in class at least gets 20% engagement, vs 0%. I think there's some value, for them, in replacing some text keywords with a graphical layout. But... Pd and Max suggest that you have access to the power of programming, without programming, while in reality, both environments are as fiddly and nitpicky as any programming language.

VCV is brilliant but there's some fundamental way that I don't get it... I want to use it like SC or Pd but it expresses logic differently.

LLMs are moving toward the point of stating a functional spec and getting code. When that finally really works, it would be a game changer.

hjh

I like all of them for different reasons. I like Pd the most for music because it is simple and I feel produces the best sound.

I do like Touch Designer too, because I feel like you can get really advanced visual imagery very fast, compared to Gem, Jitter or Processing.

I like the idea of Open Music and I have gotten some insanely cool results with it, but it just never sounded good to me and seems tough to integrate with Pd or Max. Maybe I'll look at it again, I feel like it has potential.

I like all of them for different reasons. I like Pd the most for music because it is simple and I feel produces the best sound.

I'm biased, but I think SuperCollider has the edge here. A particularly useful feature is multichannel expansion, e.g., here's 15 detuned saw waves: Saw.ar(freq * Array.fill(15, { detun ** Rand(-1, 1) })) assuming input parameters named freq and detun. The array of detuning factors automatically expands up to the * for frequency, and further up to the oscillator itself. (Max's mc. objects help with this somewhat -- and Pd will get something like that soon. That will close the gap to some degree.)

Pd has oversampling, which SC doesn't, but SC has a full suite of bandlimited (or low-aliasing) oscillators built-in, and a rich set of filters (including recently some analog-modeling filters ported from various sources, which are tastily filthy), built-in granulators, etc etc... get you where you want to go, just faster and with less rebuilding of basic needs.

hjh

I have never learned Super Collider, although I would really like to, so I can't speak on any comparison, but thanks for your input. Now I want to learn it even more.

@ddw_music
In the else library or Automatism theres some really good ones, but I haven't found any in Vanilla.

I wish I had found ELSE early on -- if I had, I might have gotten a much better impression. (It even has a copy of my favorite SC resonator, Ringz.ar <--> [resonant~].)

For fun, here's a 1-minute clip from my last show, a duo techno improvisation with my friend Neil on Digitakt (all drums are his) and myself live-coding SC (with the help of a custom dialect for patterns). I'm especially happy with the granular piano washes (the source notes come from VSTPlugin, which you also have in Pd as vstplugin~; granular processing + delays + reverb are all SC). 23-0324-hoodoo-jam-shorter-clip.mp3

@60hz

Don't forget to try https://ceammc.github.io/pd-help/,

Oh, I see, a complete alternate distribution. I had gotten discouraged when I tried to add the CEAMMC externals and found that (at that time) they didn't release them for Linux... but the alternate Pd does have a Linux version.

However, I guess I would have to build it myself, because I've adjusted some things so that the UI drives me slightly less batty. In particular, I loathe the way that dragging an object to reposition it enters edit mode -- but MSP feels that it's more important to be able to duplicate-move-type. Fine, whatever, he's wrong about that(+) but at least when I build from source I can fix it so that it works properly.

(+) Should be a user preference, but it isn't... if that weird UI decision is going to be imposed on everyone by force, then I reserve the right to be grumpy about it.

Anyway, the issue is not entirely about feature availability. It's also about workflow.

hjh

In particular, I loathe the way that dragging an object to reposition it enters edit mode -- but MSP feels that it's more important to be able to duplicate-move-type.

I totally agree with you... I build pd myself too in order to desactivate this edit mode function for students and they thanks me! It's possible to use the last ceammc lib 0.9.6 from denken but... not for linux unfortunately. 99% of my students use mac or Windows so that's ok but I agree that asking for a Linux build would be a good idea.

Otherwise for audio only... there is PlugData! But I miss all nice and well organized objects grouped by their function proposed by ceammc lib.

Otherwise for audio only... there is PlugData!

It's great! But I need GEM for class...

hjh