How to Build a Bank of SoundFonts
What you need to get started:
A soundcard that plays soundfonts. The best choices here are the sblive soundcards (any) and the Emu APS
Vienna 2.3 (Free on the Creative Labs site)
A decent Sample Editor (I use Sound Forge 4.5, but anything that has looping tools will be fine)
A MIDI keyboard with velocity and aftertouch. (Aftertouch is important).
Make your .WAV files
The quality of your .WAV files are the building blocks of your SoundFont. Record them as cleanly as possible. Record as hot as you possibly can without distortion, and delete anything that has burst through the roof. These should be in standard .WAV format. Be careful to name them with names you will remember. Its also helpful during the programming process if you put a code in the name to help you find it in the sample list when making instruments. Codes are just simple abbreviations, and you can use whatever you want, but they should be short. For example, your drums sounds could be named
When you open the list you will see the samples in alphabetical order. You will save minutes with each instrument you make scrolling through the sample list. If you are planning a SoundFont of 256 instruments and 256 presets, this will translate into many hours saved.
Make sure all your wave files have their gain optimized. I recommend using a compression plugin or utility to give all samples a uniform character. They should not have any 'overs' or clipping, as you will hear these artifacts in any instrument you use it in, but you should get as close as you can to maximize gain. Some may disagree with this as the cumulative effect of combining hot samples can easily overload the output of the onboard synth. However, I feel you can adjust the level of the sample later as you build instruments and presets, and as long as you do this you can avoid clipping the output and get the best signal to noise ratio.
Tuning is probably the biggest problem with SoundFonts. Get it right at the sample level. Otherwise you will be retuning all your instruments that use the ill-tuned sample. While the sf2 format allows for micro adjustments to tuning you will get pissed when you find you always have to retune a certain sample 15-30 times at the instrument level because you neglected to do it at the beginning. Always keep a synth fired up as reference for tuning. A typical piano patch works well.
Loop nearly all your samples. Even select drums. You can use these as great effects later on in the building of great presets. The sf2 format allows you to turn off loops at other levels of the programs. You can often save time by having a looping session and do them all at once. See my looping tips.
Finally IMPORT all your samples within Vienna. You can import more than one sample at a time. If you put all your samples in one directory (recommended) when the import window opens to browse your hard drive you can rubberband about twenty samples and import them all at once. Over 20 doesn't seem to work.
Go to the menu and select "new instrument". Give it a unique name. The "New Zone: box will open and you can select any sample you want. You can select more than one sample as long as they are next to each other in the list. See? Aren't you glad you followed my advice in Step One? All you synths are together, all your drums are together, making selection very easy.
Now you have to edit the characteristics of each zone. That is, set the Range and Root Key. That's really pretty easy if you use the right click context menus and the mouse to do these. Its as easy as any sampler I've worked with and you can see the results on the big screen.
Tip: Make sure you have enough samples to cover the entire keyboard. Perhaps the only troublesome part of the sf2 format is that samples do not stretch as far as they do on a professional sampler. For example, if your sample has a root note of C3, and you try to stretch it to cover the keyboard, you will find that the sample will detune at and play the same out of tune pitch at about A6 and higher. The trick, of course, is to have a C5 version of the same sample, which will make it to the end of an 88 key keyboard. However, and here's a point most inexperienced Fontists miss, if you end up transposing and instrument down in the preset level, you may still run out of sample on the high end even though it plays perfect at the instrument level. Rule of thumb, always give yourself an extra octave of playable notes than you need. Some might say an octave and a fifth. Of course for high quality instrument emulations, you may need a sample every 4-12 notes, depending on the instrument you are emulating.
First create and program a Global Zone for the instrument. Right click the instrument and choose global zone. These are default values for all the sample zones for filter, resonance, envelopes, LFOs, etc. You can change them for each individual zone if you want later. Many times however, defining just the global zone may be enough, and you can always delete it later if you end up tweaking all the parameters of each individual zone.
At this point you should really begin to appreciate the tremendous flexibility of the sf2 format. You can stack up zones within and instrument as deep as you want. However, I recommend, unless you really know what you want, to save that for the preset building stage where you can stack instruments up. It is best to put 1-3 layers of zones in an instrument. The fewer layers of zones, the more potential use you will get out of the instrument at the preset stage. Also, do not set the FX at the instrument level. This is better done at the preset level too. Of course, there are exceptions. If you want Zone 1 and 2 to have a chorus and zone 3 to have chorus and reverb, then yes, you could define it here. But overall, its better to consider the instruments as your raw building blocks for the presets.
You build presets using the same methods as building instruments. Except this time you are shooting for the finished, polished sound. If you have 200 instruments in your SoundFont made, you probably have enough to make well over 600 unique sounding presets. . First set the global zone. Here you can set your FX send levels for the preset. Then you can click each instrument with the preset and set the 'Initial Attenuation" parameter. This basically allows you to mix the volumes of the different sounds in a tasteful way. Parameters that I advise tweaking. You can also set Coarse Tuning and Fine tuning to taste. I advise not programming filter cutoff or resonance at this level as the results are not always predictable--this is better done when programming the instrument. Though sometimes you can get the effect you want playing with them. You will find that the parameters work by multiplying the value that is defined in the instrument. Oddly Vienna may give you 127 steps for this and it may create errors. For example, if the filter is set to 1.0 (khz) and you multiply that by 127, you will have an unusable result. Enter sample values with the keyboard like 1.1, .9, .85, 1.15 and you will get useful results for most of the parameters.
Organize your presets in a way that makes sense. Generally, the way most synth manufacturers organize their presets is as follows:
Pianos and Keys
Chimes and Bells
Of course its your bank, do what you want. Personally, I like putting all my drums first, then Basses, then keys. Be creative.
To reorder presets you can use the SoundFont Bank Utility by Emu. This used to be available at their website. Haven't seen it lately though. Its tricky to work with and the docs are sparse, but I can't go into that here.
I don't have time to get into the tricks to making a good drum kit here (perhaps that will come), but I will say that the kits one makes in a SoundFont can be as interesting and as good sounding as those one can make in a professional sampler. There is no limit to the number of zones you can have in the drum preset, so you can assign 1,2 3 or more zones per key--each with their own filter setting if you want. OK one more tip: You can layer your kits which are defined at the instrument level when you make a preset. That is you can have two contrasting kits playing on top of one another. Pan them L/R. Or Use the same kit and detune by a fifth. Or a second. Its absolutely mindblowing, even for a TweakMeister!
The sound quality of well made sf2 files is just as good as any 16 bit sampler, The soundcard itself may add a very thin veil of distortion, but it is so small it will not be evident on boards that most home studio folks can afford. If you use a digital out on your soundcard, the playing field with professional 16 bit samplers is almost level. 5 years ago you would have to pay 10's of thousands to get the flexibility you have with a cheap soundcard. That's the secret the pros won't tell you. With a little creativeness, you can make the same incredible sounds they do.
Experimentation has never been easier. The layering, mixing, timestretching, panning, morphing, vectoring capabilities are easily had by anyone who works with the sf2 format. Add some cool plugins and resample in a digital audio sequencer and you are actually far beyond what a stand alone hardware sampler can do! A final tip: its not going to be long before excellent conversion tools will be in place to import SoundFonts into samplers. The Tweak is about to beta test one. This will clearly raise the bar in commercially available soundfonts. It could easily become the de facto standard that links all samplers together in a consistent format. As time goes on, we will see the hardware sampler as an extension of the SoundFont protocol, much the way laptop computers dock to their desktop hosts. Within a year or two, this will be reality. Why? you ask? Its just easier to work this way!
take from TweakHeadz Lab