-
I am trying to make an organ sound font with custom releases. My test file is attached which I invite you to download and test yourself.
The problem I have is that there is a significant and unacceptable glitch when a note is released. I have already tried everything I can think of to cure this but the problem persists.
The problem appears to be that there is no control of the volume rise time when the release occurs - using this could make a cross-fade at the critical time. However I cannot find any way to achieve this.
All comments welcome - especially any that lead to a solution.
csw900testBourdon.sf3 -
Use the original full samples, you don't need to cut the release time and use it as custom-release. just make sure the loop exists.
For the original sample's loop, select the red-backed loop symbol.
When you release the key, the rest of the loop will play as a release.
where you specify a value of 3 seconds or higher in the vol-env-release part so that the release parameter does not automatically fade out and the original ending can be heard. // may require trial and error.
The problem here is that it may take some time to exit the loop depending on the key-press time, or the release time may be shorter than the others if the end of the loop is reached.
I added a demo soundfont. It contains a single sample and this sample is used between A3 and D2.Release-Demo.sf2 -
Thanks, your method is new to me - why did I bother to read the Polyphone tutorial?
I will do some experiments with your method. I have tried out your sample and done a brief test with my own samples and it appears to work with no glitches.
Your method has many advantages. It will more than halve the time to make a sound font.
You did mention a problem which I do not yet fully understand - plus it is possible with short key presses the release will occur before the loop has even been entered - will the custom release still play? Also the full length of the sample always has to be used whatever length it is.
csw900 -
if the key-off occurs somewhere at the beginning of the loop, it will play the remaining loop part and enter the release part after that.
if the key-off happens somewhere at the end of the loop, it will immediately enter the release part.
This can be problematic for very long loops. In short loops, the problem will not be too obvious.
I tried to explain the situation with a not very successful text representation below
loop release -|---------|--- ^ -------|--- key off ^|--- key off -
Thanks for the clarification - if you are correct then it completely ruins the method for organ sound fonts. Decent pipe organ samples are normally between 3 to 5 seconds long and the releases are usually less than 2 seconds.
I have updated my demo sound font using your method and it sounds OK. However it is very difficult to tell the difference between a 2 second fade out within or before the loop and a 2 second real release at the end of the sample. If the release occurs near the start of the sample then the sound definitely does NOT continue to the end of the loop and then play the real release. It would be easy to notice this.
Clearly I will need to do some experiments to find out what is really happening!!
PS I have just looked at the waveforms on my audio editor. I appears that the release just causes the sound to fade out whenever it occurs. The real release is rarely if ever played (I was playing a note where the release waveform is clearly different to a fade waveform.)
csw900 -
i recorded the loop+release; this is what happens;Loop+Release.png
-
To bottrop,
Please describe what you have recorded in detail and then say what you think it proves. At present, although you have annotated your picture it is unclear what it is supposed to be showing/proving.
csw900 -
follow the example of orig-rel in the demo I sent you. and type a high value for release to avoid fade-out, for example:10s.
Looped-rel releases entirely using the loop part. And that's what it does with a fade out. I put it to show the difference between the two. -
Message from ziyametedemircan on@bottrop
loop-release uses only the loop part for release and doesn't care after the loop. You should try orig-rel. -
@ziya......
Just tried your suggestion - with 10 second volume release. Release results vary depending on duration on "note on" but they mostly have an incomplete fade release followed by an attenuated real release. Interesting, but definitely of no use for an organ sound font.
Occasionally with a release following several loops there is a sudden stop (i.e. no fade) immediately followed by an attenuated real release. This is what I would hope would always happen but the attenuation of the real release ruins it.
Can you explain these results - its beyond me!!
csw900 -
the picture shows that after the loop end firstly the Volume Envelope Release is performed and after that the part after the loop end at a very low volume level.
the picture proves that i am a pretty skilful draftsman.Orga.sf2 -
@bottrop
Yes you are a good draughtsman but unfortunately your file proves that Polyphone does not perform the release in the way it should. This is probably a bug because I doubt very much whether this mode of operation has ever been tested before. If you set the volume release to 100s the sound dies away much faster than you would expect.
Do we know who is in charge of Polyphone? - it seems to me that whoever it is has gone away on a VERY long holiday (or died of Covid)
PS I have just tried your soundfont in Viena64 and the release NEVER sounds in sample+release.preset.
csw900 -
Vienna's release is broken. I mentioned it before on the Vienna forum.
-
Message from ziyametedemircan on
"Can you explain these results - its beyond me!!"
The release parameter is always set to fade-out the sound. To prevent this in a red-tailed loop, it is necessary to give the release part a higher-second value (usually 6 and higher multiples of the tail's value). But even if you give it 100 seconds, what will play there is as long as the tail. When the key-off occurs, it cannot immediately switch to the tail part. First, it has to play the rest of the loop in order to reach the tail.
But if you are not using the red-tailed loop, the normal loop's release will be enough for you. because it will use the loop and will fade-out in a setting as long as you give. When the key-off occurs, it immediately starts to apply the "release" process. Because it continues to use the loop already for the release process. -
@ziya...
"When the key-off occurs, it cannot immediately switch to the tail part. " Why not? surely this is exactly what it should be doing in order to correctly play the release transient.
Is this problem due to something in the .sf2 specification? (A very old but useful document)
I appreciate that when a release occurs a note will be playing and this will have to immediately stop and the release sound will have to start playing. This phase and amplitude discontinuity will always create a glitch. Thus a crossfade of perhaps 100 to 200ms is needed. Why can this not be done?
At present it is never possible to play release transients without a glitch so the whole red-tail feature is unusable. Surely someone ought to do something about this!! Who is this person?
Is it a problem in fluidsynth?
csw900 -
i created an ugly waveform with an ugly loop to make clear what happens.
at Instrument stage i assigned the red tail loop and 0,1 second Volume Envelope Release.
at Preset stage i multiplied the Volume Envelope Release by 50 with a slightly better result.
take look and a listenugly.pngUgly1.sf2 -
@bottrop and ziya...
Thanks for your efforts but unfortunately they do not do what is required for an organ release.
I have also been experimenting with the Polyphone Tutorial method (which I think is the only method with the potential for doing what is actually needed - the only problem is the glitch when the release begins). This glitch is due to a discontinuity between the "continuous looped wave on" termination and the release beginning. A cross fade will cure the glitch but there is no way within Polyphone to "fade in" the release. (If there is a way please tell me).
Thus I have resorted to my audio editor to make the release waveform. I deleted the continuous tone from the release and added a 200ms ""fade in" to the start of the release. This - together with a 100ms fade out of the main waveform eliminates the glitch. Fade in and out times are probably not optimum but they do work well enough to prove the point.
I have added my results in the "Tutorial" preset of Orga.sf2, now called xOrga.sf2 and attached it hereto.
csw900xOrga.sf2 -
i created an ugly waveform with an ugly loop to make clear what happens.
@bottrop: It's a really good example and explains very well how the parameters work.
@csw900: It's not the polyphone's fault that the sf2 parameters work this way. You will use this soundfont with other soundfont players, and those synths will apply the parameters in the same way. Although there may be some minor interpretation differences between them, they will not change much. To implement what you want, x-fade and the like, all soundfont synth makers need to create a new specification and soundfont players must follow suit.Thanks for your efforts but unfortunately they do not do what is required for an organ release.
I know how organ releases should be. I also have my own organ soundfonts. I preferred to use the "looped-release" in these. Because when using it, it is necessary to add a lot of reverbation and the tail (release) part doesn't seem that important to me. But for others it may be really important. I've read a lot about this.
I'm just trying to answer your questions here as much as I know, but I'm not directing you. Otherwise I would have said "use looped-release" at the very beginning of the topic. By not doing that, I think I also tried to tell you how some sf2 parameters are interpreted.
Please excuse me if I'm being rude, that's not my intention. -
@ziya
I have learned a lot from this exercise so time has not been wasted.
Cross fading the looped releases works really well. My audio editor has an adjustable "Fade In" option and takes only a few moments to do. 100ms fade in is about right combined with 100ms fadeout. I now know that I am making my sound fonts by the best method and will get the highest possible sound quality.
You say you are adding a lot of reverb - You can also add reverb to proper releases - but if you do have good samples and get the releases right you do not need much, if any, reverb. A pipe organ always sounds better without artificial reverb.
Have you tried my organ eplayOrgan? - download and use it completely free from my website. It comes with some sound fonts but I now have the capability to make better ones myself.
csw900 -
I had tried ePlayOrgan before. It could also play Miditzer and some rolls as far as I can remember. But it requires a lot of tuning and I really don't want to mess with xml files. But I can definitely say that it is one of the best software for those who want to use it with a keyboard. Soundfonts usually interest me, as my work is usually related to sheet music, arrangement, composition and sequencer works.
I'm also a little apprehensive about using other soundfonts. More precisely, I want to have full knowledge of the thing I use and its construction. I have tried almost all the soundfonts on the market and found that almost 80% of them were either not properly tested or made incorrectly. Some of them did not have their attenuations set properly, some had problems with their parameters, and in others everything was messy etc.
Since I didn't want to specialize in detail on every soundont and didn't have time for it, I started making it myself.
At first there was Vienna (it only works with the Creative sound cards), then Viena came out and then Polyphone.
Especially the Polyphone is a godsend for those who do this stuff. And for those who can take the time to learn, I recommend making their own soundfonts. Whether what they're doing works or not, at least they'll learn a lot by working on it.
As for reverb: Not for applying reverb to samples or soundfont. But we need reverb to emulate ambient reverberation and echo in the final software we are working with (eg: Large Hall or Cathedral emulation). Otherwise, we will be left with some dry pipe sounds. This is what I meant.
Sign in or register to take part in discussions.
Polyphone needs you!
Polyphone is free but there are costs associated with its website and development. A small donation will help a lot.
Donate
Learn the basics
Try a tutorial
Scroll to
top
top