What I don't know about soundfonts

[93143]

I understand that. But the release sample is on a separate modulator set (isn't it? I haven't used Vienna in a while); what I was saying is that it's possible to approximate the "natural" decay of the main sample with the envelope on the release sample, even if the main sample doesn't have an envelope. I HAVE done this in Emulator X, and it works fine unless you absolutely must have the release volume track the note volume exactly.

The only other way to do it, I think, would be to have some kind of amplitude detector that can be routed between voices - does a sampler with this capability even exist? (Don't tell me... Gigastudio?)

[audiojunkie]

I agree! I think it would be wonderful to see a site like Hammersound was in its prime.

--Sean

[SampleSmith]

Whew! Thanks for the ideas. I think I will float this past Rene and see what comes of it.

However, Sacha's comments are a splash of cold water and reality. Unless one wants to make the sf2 format a raison d'être, it seems sensible to leave the noise where it is and move on to something a little less generic. Actually, it was just this noise quotient that caused me to believe there might be room in the scene for a better guitar font, but it does seem that the format is pretty well stained. I do wonder, though, if sfz is sufficiently far removed to get away from the cacophony (at least from a commercial point of view). Perhaps in it's Dimension Pro form?

It makes me wonder if Kontakt 2 makes the most sense as a place to do the development work, especially for a 'non-hacker' like me, and convert sideways and downward from there. Does that make sense?

k

[SampleSmith]

93, this would require an envelope on the release sample that would cover a time at least as long as the note sample, right?

I don't see a way in Vienna to create an envelope for the release sample that exceeds the length of the sample itself. Am I missing the boat on something?

thanks,
Keith

[Sascha Franck]

I do wonder, though, if sfz is sufficiently far removed to get away from the cacophony (at least from a commercial point of view). Perhaps in it's Dimension Pro form?

Maybe, maybe not. Dimension Pro *might* be a good start, but you gotta watch how it turns out in the long run (in terms of success).

Also, lets not forget there's no serious editor for sfz yet, let alone a crossplatform one.
Sure, editing the scripts might be less of a hassle than one might think at first - but it simply doesn't work in realtime. You need to save and reload all the time, whereas other software samplers respond immediately, even while running the patch in your host.

It makes me wonder if Kontakt 2 makes the most sense as a place to do the development work, especially for a 'non-hacker' like me, and convert sideways and downward from there. Does that make sense?

Not sure again. As far as the status quo goes, the Kontakt 1 format would probably be a better choice.
But generally, the Kontakt format seems to be quite more widely accepted as a "pro" thing than SF2 and sfz. Whether there's justification for that is an entirely different matter.

Maybe, as an independent developer (read: without a Kontakt/HALion/Whatever based player, especially adjusted for you), your best bet would still be to offer a whole variety of formats.
Of course, every softsampler does import SF2, but you won't be able to take advance of some sampler-specific features then.

Difficult decision, really...

[marce]

Hi Sascha, sorry my ignorance, but i always hear about librarys with "sympathetic resonance" or so, specially talking about pianos. Is that possible with soundfonts and some free sampler, or not?

[Sascha Franck]

marce, you better ask learjeff regarding such things. I'm not experienced in creating soundfonts at all.
All I know is that, say, Kontakt 2 has a build in convolution reverb which is used in some of the factory patches to add to the piano sustain (there's resonances sampled, loaded into the reverb and the reverb send is opened whenever some sustain pedal message arrives). This obviously isn't possible with soundfonts.

[marce]

Thanks the explanation!

[SampleSmith]

marce, there are some IR's at noisevault, taken with the damper on and off that you might find useful
http://noisevault.com/index.php?option= ... cat&cat=17
near the bottom of the page.

I doubt it's a lot of help, since there'd be no way to adjust in the player, but might help at the sample level?

Keith

[marce]

Thanks Keith! Downloading now..
Thinking a little about that, sympathetic resonance dont mean that when you play a note with the pedal down, some other strings sound because "symphaty" with the one you press, so each note will have diffrent need diffrent impulse? Or im miss understanding?.
Coming back to the soundfont world, even if someone will sample that resonances, we would need something like a SIR plugin with midi in, i guess.
Sorry the mess, just trying to eat the info.

[SampleSmith]

Well, beng that the IR's were taken by striking the soundboard directly, I'm thinking you'd get a 'broadband' kind of resonance, if you will, so I doubt this really addresses the sympathetic part very well.

There's also the freebie 'Pianoverb'at http://www.pspaudioware.com/indexen.htm ... piano.html; that may suit better.

93143 -I seem to have that working. I'm playing with it a bit now.
thanks,
Keith

[93143]

Great - let me know if it works out.

I can't actually take credit for the silent loop idea - I found it in an old acoustic guitar soundfont on Hammersound. It generates release tick/scrape sounds on note off, which I had thought was impossible, so I got curious and investigated...

[learjeff]

SF2 has no concept of a convolution engine, so right: you'd need a SIR that takes MIDI input to know whether the damper is up or down.

Giga 3 has a convolution engine. I only have Giga 2 and no fancy pianos anyway, so I don't know whether it makes that much difference. For the forseeable future even that's a limited (though maybe effective) imitation of the real thing. With convolution you could model pedal down response for 88 keys (Sasha's "broadband") as well as one response for each single note.

But it's not practical to model every combination of up to 8 or so keys being pressed with pedal up. I suspect the fancy samplers just add the response for the one-note damper-up (key down), which would be different than actually having multiple notes down. (That's a testable hypothesis, but I'm sure not going to bother with it, and I bet that it's way beyond the ability of my ears. Frankly, I think just using SIR with "on" and "off" controlled by MIDI sustain pedal would be more than good enough for me.

Of course, multi-velocity samples for releasing the pedal would be crucial. (OK, just kidding!)

[lawapa]

What you are describing was implemented within Dimension Pro. As to how Rene did this I not sure. But it was written up as a SFZ maybe? Sympathetic resonance derived from a keypress without strike sustain pedal not engaged. Dpro also uses waveguide. Somewhat different from Convolution but similar?

Sascha; Would not Awave serve as a frontend/GUI for the SFZ format? Got it and have done sf2 conversions.

[Sascha Franck]

Sascha; Would not Awave serve as a frontend/GUI for the SFZ format? Got it and have done sf2 conversions.

Sure would. But it's not working in realtime.
See, when doing patches (in Kontakt or - still - the EXS) I often have my sequencer running with some backing track to see whether my tweakings are actually making sense. Makes things way more easier than having to edit, load, try in context, unload, re-edit, re-save, re-load, etc...
Unlesss Awave is now running as a standalone instrument as well (a la Vienna). In that case one could probably use a virtual MIDI cable to get things running (which was what I've been doing back in my SBLive days). But does it do so?