sf2 vs. DLS vs. giga (and rant)

Hey,

Does anyone have a different opinion as to the advantages of one over
the other, or can do an objective comparison?

DLS (downloadable sounds) is an open standard aimed at replacing
Soundfont, but it doesn't seem popular at all.

What are the advantages of .gig files over Soundfonts? are they just
typically larger (they sure appear to be) and more commercial based?
It seems like yet another variation on the multi-sample theme, and why
should anyone switch? Higher bit rates? more parameters?

If it's just that they are larger, sf2, which was designed for the
memory chips of soundcards, would just need a revision to be allowed
to expand into a RAM engine instrument, which for example FluidSynth
does anyway (you don't need a sf2 card to play sf2 instruments with an
engine like fluidsynth)

Carl, it seems you would know the ins and outs of this, being on the
Keyboard staff at one point...

I ask b/c LinuxSampler was designed primarily for .gig and .dls files,
and it doesn't seem to support .sf2, which apparantly the industry is
migrating away from (since thy need to make more money, they want to
convince everyone that the new boss ain't the same as the old boss is
my guess)

an example of industry fascism:

my serial port still works fine for MIDI, yet most new computers don't
have serial ports, and it's considered legacy hardware. Yet I've
*never* had latency issues (which I see happening with USB audio,
maybe not MIDI), or had to buy a MIDI converter box--I just use the PC
connectiion from my Korg X5DR rack unit, which can send 'thru' signal
to the laptop using a cheap host-to-serial cable. Works like a charm.
Yet, serial ports became USB, so people had to run out and buy USB to
MIDI converters. Then USB was bested by firewire, and firewire was
bested by USB2...the whole thing makes me laugh---as long as I can
still use my setup without paying another penny, I will.

I think that companies must plan in advance 2 format 'revolutions' or
so, and release technology they know is inferior to what is currently
possible. This gives them a buffer time to advance their R&D tech, and
that keeps the public having to buy to 'keep up'.

And then there is even money to be made on legacy die-hards---make the
legacy hardware rare and expensive, so when it needs to be replaced,
there is still a profit.

-AKJ.

>Carl, it seems you would know the ins and outs of this, being on the
>Keyboard staff at one point...

I've never heard of DLS before, but I can tell you that NO ONE
in professional music uses sf2 files. I do NOT recommend having
anything to do with GigaSampler. The codebase for that product
is totally F'd. Kontakt is the expensive sampler of choice, and
Wusikstation the cheap sampler of choice, so I'd use whatever
formats they support best. Probably they both support .gig, but
I'd have to check.

Now there may be additional issues for Linux users. That I
can't help you with.

I'm still technically a freelancer at keyboard, and you can still
see my wonderful hands every time you visit their site (the icon
for "Play" at the bottom)
http://www.keyboardmag.com

-Carl

There's also the open sfz format, which is just a text file with .wav or
.ogg files. It's what Cakewalk products like Rapture and Dimension Pro
use (which can microtune). You might be able to tune in the file itself
using region tags for each note, or specifying a nonstandard half step
in cents (I think it has to be integer though.)

http://www.cakewalk.com/DevXchange/sfz.asp
<http://www.cakewalk.com/DevXchange/sfz.asp>

--- In MakeMicroMusic@yahoogroups.com, "Aaron Krister Johnson"
<aaron@...> wrote:
>
> Hey,
>
> Does anyone have a different opinion as to the advantages of one over
> the other, or can do an objective comparison?
>
> DLS (downloadable sounds) is an open standard aimed at replacing
> Soundfont, but it doesn't seem popular at all.
>
> What are the advantages of .gig files over Soundfonts? are they just
> typically larger (they sure appear to be) and more commercial based?
> It seems like yet another variation on the multi-sample theme, and why
> should anyone switch? Higher bit rates? more parameters?
>
> If it's just that they are larger, sf2, which was designed for the
> memory chips of soundcards, would just need a revision to be allowed
> to expand into a RAM engine instrument, which for example FluidSynth
> does anyway (you don't need a sf2 card to play sf2 instruments with an
> engine like fluidsynth)
>
> Carl, it seems you would know the ins and outs of this, being on the
> Keyboard staff at one point...
>
> I ask b/c LinuxSampler was designed primarily for .gig and .dls files,
> and it doesn't seem to support .sf2, which apparantly the industry is
> migrating away from (since thy need to make more money, they want to
> convince everyone that the new boss ain't the same as the old boss is
> my guess)
>
> an example of industry fascism:
>
> my serial port still works fine for MIDI, yet most new computers don't
> have serial ports, and it's considered legacy hardware. Yet I've
> *never* had latency issues (which I see happening with USB audio,
> maybe not MIDI), or had to buy a MIDI converter box--I just use the PC
> connectiion from my Korg X5DR rack unit, which can send 'thru' signal
> to the laptop using a cheap host-to-serial cable. Works like a charm.
> Yet, serial ports became USB, so people had to run out and buy USB to
> MIDI converters. Then USB was bested by firewire, and firewire was
> bested by USB2...the whole thing makes me laugh---as long as I can
> still use my setup without paying another penny, I will.
>
> I think that companies must plan in advance 2 format 'revolutions' or
> so, and release technology they know is inferior to what is currently
> possible. This gives them a buffer time to advance their R&D tech, and
> that keeps the public having to buy to 'keep up'.
>
> And then there is even money to be made on legacy die-hards---make the
> legacy hardware rare and expensive, so when it needs to be replaced,
> there is still a profit.
>
> -AKJ.
>

[Non-text portions of this message have been removed]

--- In MakeMicroMusic@yahoogroups.com, Carl Lumma <carl@...> wrote:
>
> >Carl, it seems you would know the ins and outs of this, being on the
> >Keyboard staff at one point...
>
> I've never heard of DLS before, but I can tell you that NO ONE
> in professional music uses sf2 files.

i use soundfonts, and so does jeff harrington. probably more do too. i
guess we count as 'no one'.

here's a question--we know some soundfonts are bad...there are some
good ones though (early patches)...are there any lame .gig files or
kontakt samples?

if you are right in what you say, it surely has more to do with
lemming phenom than with any objective virtue--after all, 90% of
people use windows, and we know it's terrible.

> I do NOT recommend having
> anything to do with GigaSampler. The codebase for that product
> is totally F'd. Kontakt is the expensive sampler of choice, and
> Wusikstation the cheap sampler of choice, so I'd use whatever
> formats they support best. Probably they both support .gig, but
> I'd have to check.
>
> Now there may be additional issues for Linux users. That I
> can't help you with.

i think kontakt runs in vst on linux.

gig files work in linux sampler.

At 10:50 AM 12/6/2007, you wrote:
>--- In MakeMicroMusic@yahoogroups.com, Carl Lumma <carl@...> wrote:
>>
>> >Carl, it seems you would know the ins and outs of this, being on the
>> >Keyboard staff at one point...
>>
>> I've never heard of DLS before, but I can tell you that NO ONE
>> in professional music uses sf2 files.
>
>i use soundfonts, and so does jeff harrington. probably more do too. i
>guess we count as 'no one'.
>
>here's a question--we know some soundfonts are bad...there are some
>good ones though (early patches)...are there any lame .gig files or
>kontakt samples?

Probably.

>if you are right in what you say, it surely has more to do with
>lemming phenom than with any objective virtue--after all, 90% of
>people use windows, and we know it's terrible.

Nice troll. I work at Apple and I think Windows is good. Anyway,
no, there's nothing wrong with sf2 that I know of, although the
pro formats do I think support more information about the samples,
which the programs use when rendering them. Not sure about this,
actually, but I'd bet you an ice cream cone.

>> Now there may be additional issues for Linux users. That I
>> can't help you with.
>
>i think kontakt runs in vst on linux.

Not without breaking the law, probably.

-Carl

--- In MakeMicroMusic@yahoogroups.com, Carl Lumma <carl@...> wrote:
>
> At 10:50 AM 12/6/2007, you wrote:
> >--- In MakeMicroMusic@yahoogroups.com, Carl Lumma <carl@> wrote:
> >>
> >> >Carl, it seems you would know the ins and outs of this, being on the
> >> >Keyboard staff at one point...
> >>
> >> I've never heard of DLS before, but I can tell you that NO ONE
> >> in professional music uses sf2 files.
> >
> >i use soundfonts, and so does jeff harrington. probably more do too. i
> >guess we count as 'no one'.
> >
> >here's a question--we know some soundfonts are bad...there are some
> >good ones though (early patches)...are there any lame .gig files or
> >kontakt samples?
>
> Probably.
>
> >if you are right in what you say, it surely has more to do with
> >lemming phenom than with any objective virtue--after all, 90% of
> >people use windows, and we know it's terrible.
>
> Nice troll. I work at Apple and I think Windows is good. Anyway,
> no, there's nothing wrong with sf2 that I know of, although the
> pro formats do I think support more information about the samples,
> which the programs use when rendering them. Not sure about this,
> actually, but I'd bet you an ice cream cone.

What flavor? :)

What I do know is that Garriton's new Cello sampler corrects phase
transitions between samples to give clean portamento and legato
behavior, which adds to the remarkable realism. Some of the best
sampling technology I've heard. Garriton also seems to have some
sympathy for the Linux crowd, which is a major plus.

Anyway, I have nothing against Apple, except it's
cutesy-yuppie-crunchy 'VW Bug' image, and I wish OS-X, from a
development point-of-view, went all the way in it's unix philosophy,
and allowed developers to access hardware as file devices, like on
*nix...the whole proprietary layer of CoreAudio(tm), etc. is just
lame. I like to tinker and hack, and, well, Apple and that other OS
just suck as far as that goes. Any system that doesn't allow you to do
this:

cat /dev/urandom > /dev/dsp

to hear white noise, is just lame, lame, lame. The above paradigm is
pure beauty, and give the power user some flexibility and, well, power.

I realize most users don't want to develop custom programs, so this
probably doesn't matter all that much to them.

Windows bites. Even Windows users know this. I've seen more lousy
buggy behavior on Windows, particularly on laptops. My mother in law
is one her third windows laptop in 4 years, I think. And, my god, talk
about an ugly, icky-corporate mess of an API. Again, to get device
files, one would have to use that awful unix-simulation layer, which
last I check, sucked.

> >> Now there may be additional issues for Linux users. That I
> >> can't help you with.
> >
> >i think kontakt runs in vst on linux.
>
> Not without breaking the law, probably.

Why should they care as long as one pays for it, what system it runs on?

-A

>What I do know is that Garriton's new Cello sampler corrects phase
>transitions between samples to give clean portamento and legato
>behavior, which adds to the remarkable realism. Some of the best
>sampling technology I've heard. Garriton also seems to have some
>sympathy for the Linux crowd, which is a major plus.

Garritan is good stuff. If I remember correctly it was (or used
to be) built on the Kontakt engine. But you know, I don't actually
know all that much about samplers.

>Anyway, I have nothing against Apple, except it's
>cutesy-yuppie-crunchy 'VW Bug' image, and I wish OS-X, from a
>development point-of-view, went all the way in it's unix philosophy,
>and allowed developers to access hardware as file devices, like on
>*nix...

10.5 is officially unix (whatever that means... full POSIX complaince
or something like that). But it has had file abstractions of many if
not all hardware since 2001. I address serial ports and disk volumes
as files all the time.

>the whole proprietary layer of CoreAudio(tm), etc. is just
>lame.

Actually, having a dedicated API like that allows much faster
transactions than going through the filesystem.

>Windows bites. Even Windows users know this. I've seen more lousy
>buggy behavior on Windows, particularly on laptops. My mother in law
>is one her third windows laptop in 4 years, I think. And, my god, talk
>about an ugly, icky-corporate mess of an API. Again, to get device
>files, one would have to use that awful unix-simulation layer, which
>last I check, sucked.

OK. I'm not going to get into it.

>> >> Now there may be additional issues for Linux users. That I
>> >> can't help you with.
>> >
>> >i think kontakt runs in vst on linux.
>>
>> Not without breaking the law, probably.
>
>Why should they care as long as one pays for it, what system it runs on?

VST isn't GPL. You need the code that brokers VST traffic, and
it's quasi-legal. That's why Ardour doesn't come with VST out of
the box.

-Carl

Aaron Krister Johnson wrote:

> If it's just that they are larger, sf2, which was designed for the
> memory chips of soundcards, would just need a revision to be allowed
> to expand into a RAM engine instrument, which for example FluidSynth
> does anyway (you don't need a sf2 card to play sf2 instruments with an
> engine like fluidsynth)

I did look at Sound Fonts a while back when I was planning a program to add tuning data to them. I remember some limitation in the number of keyboard zones you can have. Probably a 16 bit number used for the zone number. It was a problem for what I wanted because a full GM sample set with one zone per note (to set a unique tuning) would blow it up. The fix wouldn't be backwards compatible, so it'd essentially be a new format.

I don't know about the other ones you mentioned.

BTW, anybody know if FluidSynth is retunable from Csound?

Graham

--- In MakeMicroMusic@yahoogroups.com, "Joe" <tamahome02000@...>
wrote:
>
>
> There's also the open sfz format, which is just a text file
with .wav or
> .ogg files. It's what Cakewalk products like Rapture and Dimension
Pro
> use (which can microtune). You might be able to tune in the file
itself
> using region tags for each note, or specifying a nonstandard half
step
> in cents (I think it has to be integer though.)
>
> http://www.cakewalk.com/DevXchange/sfz.asp
> <http://www.cakewalk.com/DevXchange/sfz.asp>
>
>
> --- In MakeMicroMusic@yahoogroups.com, "Aaron Krister Johnson"
> <aaron@> wrote:
> >
> > Hey,
> >
> > Does anyone have a different opinion as to the advantages of one
over
> > the other, or can do an objective comparison?
> >
> > DLS (downloadable sounds) is an open standard aimed at replacing
> > Soundfont, but it doesn't seem popular at all.
> >
> > What are the advantages of .gig files over Soundfonts? are they
just
> > typically larger (they sure appear to be) and more commercial
based?
> > It seems like yet another variation on the multi-sample theme,
and why
> > should anyone switch? Higher bit rates? more parameters?
> >
> > If it's just that they are larger, sf2, which was designed for the
> > memory chips of soundcards, would just need a revision to be
allowed
> > to expand into a RAM engine instrument, which for example
FluidSynth
> > does anyway (you don't need a sf2 card to play sf2 instruments
with an
> > engine like fluidsynth)
> >
> > Carl, it seems you would know the ins and outs of this, being on
the
> > Keyboard staff at one point...
> >
> > I ask b/c LinuxSampler was designed primarily for .gig and .dls
files,
> > and it doesn't seem to support .sf2, which apparantly the
industry is
> > migrating away from (since thy need to make more money, they want
to
> > convince everyone that the new boss ain't the same as the old
boss is
> > my guess)
> >
> > an example of industry fascism:
> >
> > my serial port still works fine for MIDI, yet most new computers
don't
> > have serial ports, and it's considered legacy hardware. Yet I've
> > *never* had latency issues (which I see happening with USB audio,
> > maybe not MIDI), or had to buy a MIDI converter box--I just use
the PC
> > connectiion from my Korg X5DR rack unit, which can send 'thru'
signal
> > to the laptop using a cheap host-to-serial cable. Works like a
charm.
> > Yet, serial ports became USB, so people had to run out and buy
USB to
> > MIDI converters. Then USB was bested by firewire, and firewire was
> > bested by USB2...the whole thing makes me laugh---as long as I can
> > still use my setup without paying another penny, I will.
> >
> > I think that companies must plan in advance 2
format 'revolutions' or
> > so, and release technology they know is inferior to what is
currently
> > possible. This gives them a buffer time to advance their R&D
tech, and
> > that keeps the public having to buy to 'keep up'.
> >
> > And then there is even money to be made on legacy die-hards---
make the
> > legacy hardware rare and expensive, so when it needs to be
replaced,
> > there is still a profit.
> >
> > -AKJ.
> >
>
>
>
>
> [Non-text portions of this message have been removed]
>

***Actually, I am now using the *FREE* sfz player with Sibelius,
since it is the only way I can currently get Siblius to play
quartertones accurately. As it turns out, their current softsynth
engine, Kontaktplayer2, plays them *INCORRECTLY!* So, I can't use
it. However, the sfz player plus soundfonts from, curiously
enough, "Finale Notepad" worked terrifically for me, and Sibelius is
again a microtonal music tool... at least for quartertones.

J. Pehrson

--- In MakeMicroMusic@yahoogroups.com, "Joseph Pehrson" <jpehrson@...>
wrote:

I should have mentioned the free sfz player, which also plays
soundfonts. It's supposed to have a very good playback engine.

Joe

> ***Actually, I am now using the *FREE* sfz player with Sibelius,
> since it is the only way I can currently get Siblius to play
> quartertones accurately. As it turns out, their current softsynth
> engine, Kontaktplayer2, plays them *INCORRECTLY!* So, I can't use
> it. However, the sfz player plus soundfonts from, curiously
> enough, "Finale Notepad" worked terrifically for me, and Sibelius is
> again a microtonal music tool... at least for quartertones.
>
> J. Pehrson
>

> ***Actually, I am now using the *FREE* sfz player with Sibelius,
> since it is the only way I can currently get Siblius to play
> quartertones accurately. As it turns out, their current softsynth
> engine, Kontaktplayer2, plays them *INCORRECTLY!* So, I can't use
> it. However, the sfz player plus soundfonts from, curiously
> enough, "Finale Notepad" worked terrifically for me, and Sibelius is
> again a microtonal music tool... at least for quartertones.

Joseph-

so you admit you use soundfonts? Carl will now add you to his list of
non-professionals! ;)

-A.

If I may clarify my comments, I didn't mean to offend anyone
using soundfonts. The phrase 'nobody does xyz' is idiomatic.
Soundfonts are wonderful because of the community, do-it-yourself
attitude around them (crystallized at Hammersound), but the
format has both cultural (e.g. what supports it and how) and
technical (e.g. Can they store convolution reverb impulses?)
limitations that are significant. If there's an interest in
improving the production quality of the music being made around
this list, getting some modern computer music chops is highly
recommended. I'm the most guilty party of all... the only
sample format I've ever used with my own music is (wait for it)
sf2. I have a pretty comprehensive sf2 sample library on my
hard drive, in fact. But I've also used (at work) some top-notch
software and I can attest that "there is a difference".

-Carl

At 10:18 AM 12/10/2007, you wrote:
>
>> ***Actually, I am now using the *FREE* sfz player with Sibelius,
>> since it is the only way I can currently get Siblius to play
>> quartertones accurately. As it turns out, their current softsynth
>> engine, Kontaktplayer2, plays them *INCORRECTLY!* So, I can't use
>> it. However, the sfz player plus soundfonts from, curiously
>> enough, "Finale Notepad" worked terrifically for me, and Sibelius is
>> again a microtonal music tool... at least for quartertones.
>
>Joseph-
>
>so you admit you use soundfonts? Carl will now add you to his list of
>non-professionals! ;)
>
>-A.

--- In MakeMicroMusic@yahoogroups.com, Carl Lumma <carl@...> wrote:
>
> If I may clarify my comments, I didn't mean to offend anyone
> using soundfonts. The phrase 'nobody does xyz' is idiomatic.
> Soundfonts are wonderful because of the community, do-it-yourself
> attitude around them (crystallized at Hammersound), but the
> format has both cultural (e.g. what supports it and how) and
> technical (e.g. Can they store convolution reverb impulses?)
> limitations that are significant. If there's an interest in
> improving the production quality of the music being made around
> this list, getting some modern computer music chops is highly
> recommended. I'm the most guilty party of all... the only
> sample format I've ever used with my own music is (wait for it)
> sf2. I have a pretty comprehensive sf2 sample library on my
> hard drive, in fact. But I've also used (at work) some top-notch
> software and I can attest that "there is a difference".

Hey Carl, I'm just busting your chops a little for your blanket
comment....

Anyway, I think I can say that the most important thing is not
necessarily the tools but how they are used...the proof is how we
often make technical improvements, and then often desire 'retro'
sounds for aesthetic 'cache' reasons. I think cool music can be often
made with cheap-o equipment, including legacy crappy soundcards. Of
course the music must match the aesthetic...I wouldn't want Bach this
way, but some cool microtonal self-consciously cheezy robot techno
could be great this way!

With soundfonts, I think the real limitation is often that the
distributed fonts are of low quality and realism due to trying to make
everything loop in a really short time...the highest quality fonts I
have heard have no looping or really large loops, and at the very
least, if short, painstakingly chosen short loops. Of course this
usually means that most GM fonts are out of the question--to do
anything of quality, I think one should collect together some large
custom fonts for each instrument desired, each made up of long
samples, and if necessary b/c of memory limitations, one can overdub.
I think this tend to be the difference between most sf2 files and
.giga files---raw size...so I think it's not as much a flaw of sf2 as
much as the culture of how they tend to be employed.

I'm experimenting for instance with a horn font that uses the Iowa
university anechoic samples...it's interesting how realistic one can
get with care...and there's no comparing the sounds I'm getting with
your average GM horn sound, because the GM sets, in order to pack 128
presets, save memory by looping really soon into each sample, and you
get this awfully cheap 'wavetable sound' so often...the most realism
comes with the large scale variability and 'chaos' of the performer's
air pressure, and this is usually only captured with a large loop. It
would take an ace font artist to conjure this effect with a small loop.

re: convolution impulses--what format uses them? I usually do reverb
through a plugin or 'post-production' in an editor, so I'm less
concerned about that--I tend to be quite modular with tools. DLS, by
the way, is the MIDI manufacturers new standard, but it for whatever
reason, hasn't got the mindshare that sf2 or even sfz have.

Having said all this, you areright that the newer formats and engines
are adding new innovative features that blow sf2 out of the
water---like I mentioned, the phase transitioning of Garritan's new
cello product. But this does not mean we can't still sqeeze more juice
from sf2...

On other thing I noticed deserves mention: the problem with a lot of
these commercial products is poor or nonexistant support for
microtuning. So you have great sound, but you have to jump through
hoops...

-AKJ.

>re: convolution impulses--what format uses them? I usually do reverb
>through a plugin or 'post-production' in an editor, so I'm less
>concerned about that--I tend to be quite modular with tools.

If you want to sample a room for later use, I believe you can save
the samples in .gig format. Certainly great halls and cathedrals
are sold in .gig. That's just one example of course. Graham gave
another. And I believe some formats have metadata to label transition
samples for strings and so forth. But again, I haven't done a whole
lot with this (mainly because samplers are not very interesting to
me personally).

>DLS, by
>the way, is the MIDI manufacturers new standard, but it for whatever
>reason, hasn't got the mindshare that sf2 or even sfz have.

Ah.

>Having said all this, you are right that the newer formats and engines
>are adding new innovative features that blow sf2 out of the
>water---like I mentioned, the phase transitioning of Garritan's new
>cello product. But this does not mean we can't still sqeeze more juice
>from sf2...

Sure. But there's a danger here too, at least for certain personality
types. We learn one way, and then we tend to turn to that in the
future. I'm stuck with Encore and Audio Compositor when I should be
using Sibelius and REAPER.

>On other thing I noticed deserves mention: the problem with a lot of
>these commercial products is poor or nonexistant support for
>microtuning. So you have great sound, but you have to jump through
>hoops...

Kontakt is really powerful, with its built-in scripting language.
Also you can get the 12equalboresme tools for it. Wusikstation
has Scala support.

-Carl

--- In MakeMicroMusic@yahoogroups.com, Carl Lumma <carl@...> wrote:

> samplers are not very interesting to me personally...

does this mean you are hardcore acoustic mostly? or that you tend to
like only straight classic additive/subtractive/physical-modelling
when in the electonic realm?

At 08:08 AM 12/11/2007, you wrote:
>--- In MakeMicroMusic@yahoogroups.com, Carl Lumma <carl@...> wrote:
>
>> samplers are not very interesting to me personally...
>
>does this mean you are hardcore acoustic mostly? or that you tend to
>like only straight classic additive/subtractive/physical-modelling
>when in the electonic realm?

The latter. -Carl

A couple of days ago Carl wrote:

> Wusikstation has Scala support.

It also does better than soundfonts (sf2), at least for some purposes. I
converted all of the Early patches, for example from sf2, to Wusik format,
and they seem to perform legato more realistically with wind sounds than if
you just play the sf2 files through a soundfont player.

Rick