Physical synthesis

October 30, 2001 1:05 AM

Physical synthesis models vibrating structures to synthesis sounds (Tao home). SAOL lets you program audio synthesis inside an MP4 file (MP4 home).
posted by andrew cooke (10 comments total)

Sorry that these are somewhat technical and, by internet standards, rather old, but I thought they were damn neat and wanted to share. I have no throwaway comment to provoke discussion.
posted by andrew cooke at 1:06 AM on October 30, 2001

I looked at SAOL a couple of years ago, and found that it lacked CSound's muscularity. Then again, it was, as I recall, originally intended as a low-bandwidth protocol language which might stimulate really low bandwidth music production and distribution (ie, web sound). It, like CSound, has never found the audience it deserves. In CSound's case, mainstream interest was shooed away by snooty elitists. SAOL, I suspect, was never developed for - IOW, a glam GUI was never produced, as commercial viability would have been very difficult to prove. In 1999, the target was moving so fast...

All alternative synthesis schemes suffer truly awesome competition from houses like Korg et al, who devise their own synthesis algorithms, and of course, VST, whose plugins and effects grow more golden with every jump in processor speed.

What I'm saying, andrew, is that I agree with you - this stuff is all very damn neat. I do a lot of sound programming, and I develop custom noises. But it is very difficult, and will continue to be very difficult, to have much widespread commercial impact on a very active, yet curiously traditional software/sound scene.
posted by Opus Dark at 1:59 AM on October 30, 2001

To quote from the second link:
'We assume some familiarity with computer programming and algorithms, audio signal processing, and music and sound.'
Good for all those people who always wanted a 20ft guitar, but lacked the physical resources to make one. Or something.
posted by asok at 2:01 AM on October 30, 2001

I thought CSound had got recognition. It seems to be the reference approach to synthesis in the Open Source world (I'm not sure this kind of thing will ever be mainstream). But it looks horribly messy.

I don't use a MS OS, but got the impression VST was more used for commerical plugins that for people doing their own synthesis. Do you think SAOL is dead? I was wondering about implementing something like Tao in it. Maybe PD or jMax instead?

Sorry, didn't intend this to be a thread to discuss my own personal coding plans. FWIW, Tao (the first link) actually could simulate a 20ft guitar, in some aspects...
posted by andrew cooke at 2:31 AM on October 30, 2001

Tao (the first link) actually could simulate a 20ft guitar, in some aspects

Yeah, but can you smash it onstage?
posted by Outlawyr at 6:16 AM on October 30, 2001

Csound is great, but I'd suggest checking out the Mac-only Supercollider. I'm just getting into it myself, so I'm not sure yet if it features physical modeling, but it does seem to be extensible. The results are very nice, the generative code compact and easily hackable (not something I'd say about Csound), and CPU efficiency seems reasonable.
posted by retrofut at 5:43 PM on October 30, 2001

Kudos for the poster!
*Expecting interesting input any minute now...*
posted by MiguelCardoso at 1:51 AM on October 31, 2001

OK, here's what I got from someone involved in the development of SAOL (John Lazzao). He actually said a bit more off the record, but this has the gist of it, with a few personal details removed (nothing exciting, he just didn't want to speak for others).

[Incidentally, I heard of this and first contacted John Lazzao via the linux-audio-dev mailing list]

---

[Me:] Thanks. My main question is - what's your impression of the future of SAOL et al in MP4?

A multi-dimensional question:

[1] MPEG 4 Systems mandates Structured Audio as the way for mixing together the audio from multiple codecs. So, if you think that MPEG 4 Systems will become the operating system for building multimedia systems that are more complex than a single audio or audiovisual signal, then SAOL has a future as mandated by the ISO/IEC standards
for MPEG 4.

[2] Computational audio in general has a future -- for pplications such as Network Musical Performance, long-distance audiology tests, special-purpose encoder/decoders whose algorithms are customized to
the data source, etc. These are more future applications than present applications, and its always hard to predict whether an existing mature standard like Structured Audio will be pressed into service once an application domain takes off, or whether a new language will
be written customized for it. But there's at least a chance that a somewhat reasonable, completely standardized language like SAOL will catch the eye of the folks making this decision.

In short, divining the future of SAOL is a different business than divining the future of most new programming language, because the presence of MPEG as a standardization body changes the equation significantly, especially since some of the candidate markets (telecom, cable TV, Internet infrastructure, DoD) have a history of choosing international standards if possible.

Do you know of anyone using it?

Apache logs from our website, with the usual cautions about the limits of making conclusions from log numbers:

sfront main home book home all page
distrib. page page views

mar99 66 103 0 103
apr99 17 79 0 79
may99 20 127 14 264
jun99 16 92 8 145
jul99 20 257 73 952
aug99 104 495 195 2459
sep99 686 1186 299 4644
oct99 171 459 174 2169
nov99 777 1256 334 5112
dec99 480 959 292 4236
jan00 976 1630 480 7205
feb00 1123 1918 560 8748
mar00 796 1209 438 6396
apr00 681 1376 425 5793
may00 1321 3512 802 11292
jun00 1463 4224 771 11267
jul00 1342 3935 824 13573
aug00 1469 3778 629 10366
sep00 1607 4391 607 11414
oct00 1226 4531 654 12822
nov00 1251 4132 598 20365
dec00 984 4230 596 10727
jan01 1337 4320 648 11038
feb01 1203 4052 543 11862
mar01 1126 4231 566 11175
apr01 1534 3795 544 10218
may01 969 3469 545 8834
jun01 1112 3606 584 9707
jul01 1059 3538 656 10604
aug01 1341 3560 591 10156
sep01 906 3172 497 8567
----- ----- ---- -----
total 27183 77622 13947 242292

[I tried to put that in pre, but it freaked out the Mefi format]

Will anyone respond to bug reports in sfront?

It's research software, so future support is always has a risk element to it. The source is under the GPL, though, which limits this risk. Also, the software itself is pretty mature at its core, look at the change log updates to sense this:

http://www.cs.berkeley.edu/~lazzaro/sa/sfman/user/ref/index.html

Finally, note that sfront != SAOL -- its just the most popular implementation right now, but anyone can take the spec and make their own compliant SAOL decoder.

When I looked at links and discussions there's a clear drop in interest over the last year. Is this just because the hard work is done, plus the general slump, or is it simply not being used by anyone.

I think the web log numbers, and my own rate of email about sfront this week, show a certain sustained level of interest. But I think going forward, the interest in SAOL will really rise or fall on the interest of new applications built on SAOL, using sfront or some other infrastructure -- our work on network musical performance is one example of an application which could take off, and part of the reason we've documented the APIs so well in sfront is to support other people who have their own application ideas.
posted by andrew cooke at 2:58 AM on October 31, 2001

SuperCollider does look nice, but... :-)
posted by andrew cooke at 3:04 AM on October 31, 2001

Thanks for the follow-up, andrew. v-shall-c.

Good luck with whatever noise-makers you are working on!
posted by Opus Dark at 3:28 AM on October 31, 2001