Soundfont technology out-of-date?

[useruseruser]

!!! IMPORTANT !!!
Please read this text, before you just sloppy do a vote!


Soundfont technology out-of-date?

Some thoughts to soundfonts and the tech from view of today.

When the soundfont technology was developed and given to the consumer market it was indeed a big bang. The collaboration between EMU and Creative Labs was for sure very innovative and has changed the whole world of producing MIDI based music on computer systems. This was before 15 years so far and the state of the art of computer based sampling tech at this time.

(Note: Creative were not the first, who did it. Gravis (Ultrasound) and TurtleBeach were somewhat earlier and developed the soundfont-idea in consumer computer soundcards with also concrete results...)

But in the meantime the situation has changed something. We are 15 years older now. There are also many new formats. (The native Windows format is for exsample DLS - used by the operating systems software wavetable synthesizer.)

Many people still love soundfonts and collect them with hot abandon. The quality of available soundfonts has also increased significantly. Even the free ones. (The first user soundfonts had very pure quality, because the RAM memory limit of the original SoundBlaster AWE cards, which merely had 512kb ! of inbuilt memory and the upgrade possibility to 8 or 16MB RAM).

Novadays technology removes any RAM limitation nearly unlimited. It is even possible to stream soundfonts from disk intelligently to reach highest amount of performance and quality. Thus, naturally the sound quality is better now.

But is this enought for successful evolution?

Seems, that many users (developers and creators) still live anyhow in the past of the good old soundfonts, by using them with premature technologies. Many (latest) modern soundfont players don't even try to extend the concept somehow. Some examples of the limitations of soundfonts (the specification of soundfonts also came somewhat into the years) from my point of view:


- Working with static velocity layers to achieve realistic expression and dynamic.
This is indeed a very limited and oldfashioned solution, not only because it's cumbersome to create, but also very limited in the final sonic quality. And needs allot of additional sound memory.

- Missing Velocity crossfading or key-zone crossfading therefore.

- Missing extendable realtime playing modulation possibilities (esp. for Velocity and Aftertouch).
This even limits the sonic possibilities just more significant. It is not possible to modify the necessary dynamic articulations for most natural instruments satisfactory. Even not with the latest HQ players.

- Limited implementation of intenal modulation routings and effects.
There are merely a few average effect implementations and internal routing possibilities supported by the standard. This is no longer on top of time amd mostly unpleasant.

- Real stereo sounds must be created with internal tricky layering (not directly supported by many editors).

- ... probably many things more to count here ...


The result is finally, that even today the usage of soundfonts has still a touch of semi-professional and often produces relatively unpleasant sonic results.

As good a particular soundfont may be, as static it sounds never the less on each standard soundfont player/editor. Simply due the missing features and a missing more professional layering concept.

But the basic concept has just more power in it!

The people still love the usage of soundfonts, but mostly with the result of a kind of "instant" effect, so that you actually *can hear* that they used a soundfont (this is merely my thesis by using *real instrument* soundfonts). Some rare exceptions.


What I finally want to say, is the following:

The technology for improving the soundfont technology is still there, but not used!

Fortunately, today most SoftSamplers support the direct import and editing! of all possible soundfonts. This gives the technology the right direction and the necessary sonic quality. But it is merely a "side effect", not really an extension or development of the soundfont format.

If you map and edit your soundfonts with professional grade samplers, you are actually able to remove all those limitations and add shopisticated new dynamic to the theme. Of course, then the soundfonts are no longer "backwards compatible", which means they could not be used with the same qualities on the original soundfont equipment.

So with much work, you can do great things.

Especially for good sounding orchestral sound there are for instance the following tips from me:

- Use stereo samples whereever possible for real instruments.
Even Strings and orchestral sounds will be significant more pleasing this way. Nobody really would enjoy monophonic recorded classical music, you know ? String orchestras recorded in mono simply sound strange and un-real.

- Use Velocity crossfading and advanced Velocity modulation to achieve natural sounding results with maximum of articulation dynamic.

With velo-attack, velo-release, velo-filter, velo-crossfading, key-crossfading, dynamic sample offsets, specialized resampling models and advanced Aftertouch expression you can sucessfully remove nearly all of the disadvantages of static soundfonts with very impressive results. You also may save time (for editing) and loads of users memory by doing so.

A smart mapped and filter modulated GrandPiano with 64 or 128MB RAM *can* sound with only one keyzone-layer just as expressive and dynamically as a 1GB heavy velocity layered one. The expression of nearly all solo instruments can be maximized this way enormously.

- Replace the average standard effects with high quality special effects to give the right natural ambient to the sounds.

- last but not least add your own usefull expression controllers to the patches (i.e. volume swells, filter dynamics, portamento techniques and so on) to give them live and realistic expression. This completes the intent to recreate the dynamics of a static sampled sound and moves the technology somewhat more into the "physical modelling" direction.

The possibilities with a good sampler are really exciting today.

Which particular sampler you use is not so much the question. But feature-rich it should be at least.

But can we have this tech for our soundfont players directly?

Hopefully, there will come out some more and better quality soundfont players in the future, so we can move a step further. sfz+ is still not the state of the art, I think.

Some developers still discovered the enormous potential of "instant sound generators" with an enormous feature set to modify this instant sound in realtime, but for sure they aren't interested to give you the possibility to load your own soundfonts into it for some obvious reasons (they wana make money with their sound libraries) ...



Please soundfont-technology developers: Give us more possibilities and extend this amazing technology to fit right into our time!


Also the format extension and possible implementation of compression algorithms seems to be a good idea for the future to increase the portability of soundfonts.


My thoughts end here for the moment...
Feel free to say, what you think.

[freeztar]

sfz, my friend...
SFZ+...

[useruseruser]

Sorry, but did you read the text?
The point is, that sfz+ even don't extend the idea of soundfont technology to a real professional level. It's VST a soundfont *player* with the basic legacy compatibility. Not much more.

.

[mumpcake]

Well, while it may be published, Creative owns the standard and they haven't done much with it in some time. Developers aren't going to be able to do any more with soundfont than the spec allows.

[freeztar]

Sorry, but did you read the text?
The point is, that sfz+ even don't extend the idea of soundfont technology to a real professional level. It's VST a soundfont *player* with the basic legacy compatibility. Not much more.

.

ok, fair enough...I got lost in the verbosity and wound up skimming the last half.

Having read it now...

I'm sure Bill Gates didn't look upon DOS 5 and say....hey, how can I make this better.
He said, "f**k it, I'll revolutionize".

The same goes for SF2, and midi for that matter. It just takes time for things to change.
In the future we will have complete control, internally, over velocity and all other controllers, at our fingertips. But as of now, we have to settle with the uneditable sf2.

As the point of your post is most probably to inspire developmental conversation, I would say that I agree with your notion of a paradigm shift, as far as internal manipulation goes...
For the time being, I am content with sf2 and filter effects. Nonetheless, I welcome a new breed of "all-in-one" sample players.

[VitaminD]

Sorry, but did you read the text?
The point is, that sfz+ even don't extend the idea of soundfont technology to a real professional level. It's VST a soundfont *player* with the basic legacy compatibility. Not much more.

.

ok, fair enough...I got lost in the verbosity and wound up skimming the last half.

Having read it now...

I'm sure Bill Gates didn't look upon DOS 5 and say....hey, how can I make this better.
He said, "f**k it, I'll revolutionize".

...and create DOS6!

[freeztar]

ok, it's been a looooong time...

[useruseruser]

Well, while it may be published, Creative owns the standard and they haven't done much with it in some time. Developers aren't going to be able to do any more with soundfont than the spec allows.

As far as I know, this is not entirely true. Correct me if I am wrong.

The specification is public. And the specification allows per definition the implementation of manufacturer exclusive extension chunks.

By the way. Developers are not forced to use sf2 format. There is an official format from the MIMA - MIDI Manufacturers Association too (the DLS specification, which is freely extendable).

So this one could be extended to a better standard, supporting advanced features equal to professional samplers. The problem is, that there is unfortunately no such standard. Every sampller manufacturer cooks his own soup.

The soundfont format is definitively out of time, but extremely popular still today!
So why not extend it, adding (maybe) OGG-Vorbis support and all kind of professional features now?

Sf2 can easiely be converted into DLS on the fly, which already supports compression algorithms per definition for instance.

The other way was to save an extension file nearby the standard sf2 file to allow just more professional soundfont players and editors (more modulations, more effects, more parameters and so on) without affecting the original file. Or saving the special parameter set with the synthesizers patch...

.

[Spratman]

I think the thing that people are forgetting is that, on the most part soundfonts are more than good enough for purposes such as playing back GM based compositions. For that purpose SF2's are really quite something. What they're not brilliant at is all the special articultion features. I don't think anyone has created a sampler that caters for all the possible performance nuances yet (and probably wont, much easier to release romplers/modules that do things like say play bass sounds really well).

Soundfonts provide *a basic common standard for multisamples* (not some cheap loop playback standard ala mellotron). That's part of the reason why they're so popular.

The progression I would like to see is to forget about all this proprietary nonsense and go for something with an XML inheritance based structure file format. Why do this?

- Can choose which parts of the instrument definition types you want to implement in software. Can extend the structure if you like (and publish it so that others can see what you did without huge difficulty).

- XML is supported really well these days

- People/Human readable patch files

- Doesn't matter if you're using a fruitmachine or something else

- Even a twit like me could write an editor for such patch files in Java, VB, C, C++, Delphi...

- Bitrate etc, doesn't matter. Define it in an XML file and if the player can't do it, it's upto it to decide what to do/what not to do.

- File formats could be anything from aiff to ogg to proprietary (open markup scheme, closed encrypted playback format) It's up to the software to say "I can do this".

- Could be supported by Synthedit etc quite easily

- Add whatever extra effects etc...

But yes, random XML based rant over...

Did someone say soundfont?

Spratman
That soundfont bloke..

[useruseruser]

This is indeed a very good point, I think.

OK, it may be more efficient to work with binary files in some situations, but probably not neccessarily. And probably no longer today ( the samples will be extracted and loaded into memory never the less as a chunk of data before you can play it - mostly).

But then (with your idea) we have a new format and all the thousands of sf2 files have to be converted first, before you actually can apply the new features.

This requires that the audio files will be saved into an folder structure and the XML file is part of this package, right?

So thats basicly what for instance Kontakt does, It extracts (if you save the patch) all samples from the sf2 file and creates it's own file format based on folder structures. That's indeed a solution. But only on first look.

Actually there is the situation, that you probably have 3 different patches on your disk with the ful sample set, because you want to keep the original sf2, and have two other sampler patches too. And maybe a copy of the sounds as DLS file...
Then you actually waste allot of disk memory to store one and the same patch in different folders and files on your disk...

So the ultimate standard could be highly necessary now...
I'm not that sure, what the sampler developers do think about an "universal standard" of sound patches (soundfonts).

But then again, I think, that the existing standard is already per definition extendable. But merely not used...

I don't see the problem, why such a "player soft synth" like the most popular sfz+ cannot add additional professional features (neccessary for a professional sound articulation) and save the parameter set as a usual synthesizer patch.

But probably the "ultimate new soudfont format" (maybe with inbuilt conversion features) could be the way ...

I whish I was able to minimize my time, by remapping all the good soundfonts I have with Kontakt to get out any satisfactory results finally.
I wish a had better tools to give soundfonts more sonic quality and productivity, which is simply not possible with th old "player" technology.

A soundfont is basicly not different to any other HQ sample library. Merely the playback features are frozen at the state of 1994. That's really sad.

.

[marce]

I have one word associated to the popularity of soundfont: price. I use soundfonts because i havent to pay near nothing to have medium quality/good sounds. I believe that Garritan and above are very good things, but i cant afford them. Im not alone, you know.

Beyond of this, i believe that soundblaster has dropped the development of sf2 format far time ago. See emu site: they still support the hardware based software synths, but they are not married with sf2. They load giga, and others too.

I feel this moment similar when a lot of technologies try to replace the floppy 31/4 diskettes. Zip, external drives, cdr, etc, sticks memoryes. All them better, but a lot of them are forget today. (and i still have my floppy drive). Maybe dont worst the pitty effort in develope a better soundfont format if tomorrow someones announces a Synful clone (or synful 2) that explode all libraries out there.
Meanwhile, i will keep my soundblaster live and my soundfons. Really is easy use them.
(anyone note that the filter cuttofs and modulation envelopes are supported very bad way from soundfont soft players?)

[zvon]

As a sample developper I don't think they are. We released 2 sample sets that are available in the Vsampler 3 format and as soundfonts. The Vsampler 3 format is much more versatile and powerful than soundfonts but they are a format that allows owners of other samplers to import the samples already mapped which saves them a lot of time (well at least some time)and they can then tweak them the way they want.

Of course I would like to see a more evolved standard format for samples. But for now soundfonts are the only format that can be read by almost anything.

Also, I may be wrong here, I think that a standard is by nature simpler that things that are not a standard. Another big advantage of soundfonts for those who own a soundcard that can play them directly is close to 0 latency and close to 0 CPU usage.

Don't get me wrong, if I want more control I load the samples (or the soundfonts) in a sampler, but to distribute samples they are still the best thing IMO.

[useruseruser]

As a sample developper I don't think they are. We released 2 sample sets that are available in the Vsampler 3 format and as soundfonts. The Vsampler 3 format is much more versatile and powerful than soundfonts but they are a format that allows owners of other samplers to import the samples already mapped which saves them a lot of time (well at least some time)and they can then tweak them the way they want.

Of course I would like to see a more evolved standard format for samples. But for now soundfonts are the only format that can be read by almost anything.

Also, I may be wrong here, I think that a standard is by nature simpler that things that are not a standard. Another big advantage of soundfonts for those who own a soundcard that can play them directly is close to 0 latency and close to 0 CPU usage.

Don't get me wrong, if I want more control I load the samples (or the soundfonts) in a sampler, but to distribute samples they are still the best thing IMO.

You are possibly right there.
But please think more in the sense of the users.

The key is portability and interchangebility.

At current state this is highly unsatisfactory (from the point of the users).

In my case I have multiple copies of some instruments on my harddrives, because all samplers and sample players use a different format. Thats shit.

I want to play them seemlessly on different samplers and players. So a flexible XML format could actually solve the problem for most users. By having ONE centralized sample folder and sampler-specific mapping file formats (which don't include one and the same sample copies in each format).

Kontakt is the only sampler I know from, which actually allows relative path information for the sample libraries. Thats fine, at least.

.

[useruseruser]

I have one word associated to the popularity of soundfont: price. I use soundfonts because i havent to pay near nothing to have medium quality/good sounds. I believe that Garritan and above are very good things, but i cant afford them. Im not alone, you know.

Beyond of this, i believe that soundblaster has dropped the development of sf2 format far time ago. See emu site: they still support the hardware based software synths, but they are not married with sf2. They load giga, and others too.

I feel this moment similar when a lot of technologies try to replace the floppy 31/4 diskettes. Zip, external drives, cdr, etc, sticks memoryes. All them better, but a lot of them are forget today. (and i still have my floppy drive). Maybe dont worst the pitty effort in develope a better soundfont format if tomorrow someones announces a Synful clone (or synful 2) that explode all libraries out there.
Meanwhile, i will keep my soundblaster live and my soundfons. Really is easy use them.
(anyone note that the filter cuttofs and modulation envelopes are supported very bad way from soundfont soft players?)

You are exactly right.
There are many users, who work for even that reason with soundfonts. The quality is particulary impressive up to spectacular today. You even have free soundfonts, which blow away some of the commercial libraries with bravour.

This I ment with saying: There is allot of potential due the possible technology and popularity of soundfonts. It merely has not continued to find the right way since some years.

At least sfz and sfz+ and the samplers supporting it did give it a new big kick forward in the last time.

I actually cannot understand people, who are spending their money for sample player products, which actually don't allow them to freeley exchange and load their own samples. Products, which make you to a slave. A slave of the companies, who wanna sell their (expensive) libraries to you and thereby forbid you to use the full power of their (partwise impressive) player products.

Personal opinion: Finally all those "alias stylus sampletanked fastfood" people will sound the same this way ...

However, I think, those products have no future. And that's really good that way.

And I think that universal and high quality soundfonts actually have a future. Because you are free to use and create them with whatever you want. There are no limits.

So we need some "edge cutting" developments here to keep the idea of sample freedom ...

edit: Also read the topic "sampletank as sampler". It basicly dreams an equal dream.

.

[zvon]

I want to play them seemlessly on different samplers and players. So a flexible XML format could actually solve the problem for most users. By having ONE centralized sample folder and sampler-specific mapping file formats (which don't include one and the same sample copies in each format).
.

That would be great but as each sampler has different features and sample manipulation tools it would still have to stick with the basics. But yes, it could include more than soundfonts do. The sfz format may be a step in the right direction.

Kontakt is the only sampler I know from, which actually allows relative path information for the sample libraries. Thats fine, at least.

.

Vsampler 3 and the sfz format allow that and if i'm not mistaken Halion too. I'm sure there are others.