Let's say that one day you are leafing through the pages of the latest issue of EM, and you come across an ad for a new sample library of Central Malaysian nose flutes. The trouble is that the CD is for GigaStudio, but you use SampleTank on your Mac. Can you use the sample library, and, if so, how?
The good news is that no matter what sample library and sampler you choose, you'll be able to make some sort of conversion between formats. The solution might be as simple as pointing your sampler to the source library, clicking on the import button, and entering nose-flute nirvana. Or you may experience difficulties, forcing you to leave some of the library's features behind, and you could need separate software to make the conversion (see the sidebar “Conversion Software”).
Bear in mind that converting a sample library from one sampler format to another doesn't void the library manufacturer's End User License Agreement; all the rules stated in the EULA apply to the converted library. In other words, you can convert your own library for your own use, including whatever commercial uses are permitted in the EULA.
Who Needs It?
Why should you need to worry about conversion; why not just buy the version of the library that matches your sampler? For one thing, it might not be available in your sampler's format. With the growing number of samplers on the market, it's increasingly difficult for sample-library developers to directly support multiple formats, although they do cover most of the majors.
You may own a library for a sampler that either you no longer use or is no longer made. There are a huge number of SampleCell II libraries, but most SampleCell II users have now moved on to other formats. That's why most major software samplers import SampleCell II libraries directly. (Incidentally, SampleCell I libraries are not widely supported.)
You may want to use that library in a DAW that doesn't support the software sampler plug-in for which the library is formatted. For example, if your primary sampler is GigaStudio and all your libraries are in Giga format, but you use Reason for laptop sketching, you will undoubtedly want to convert some GigaStudio instruments to Reason's NN-XT sampler format on occasion.
Finally, you may want to take advantage of the features of one sampler, while primarily using the library in another sampler. For example, Kontakt 2's Script Processor or its multiple playback modes might be just the thing to add a new twist to that overused EXS24 library that's been idly consuming hard-drive space. Software samplers are not only plentiful and varied, but they've also become very cost-effective. It's no longer a one-size-fits-all world, and that's why there will never be a universal sampler format. It's not just a matter of obstinacy or turf protecting on the part of sampler manufacturers; samplers have become as diverse as synthesizers.
FIG. 1: Reason''s NN-XT mapping editor allows zone grouping (upper versus lower), and -individual parameter settings (along the bottom) for groups and individual zones.
The only thing that most samplers have in common is that they play standard audio files in the usual sampling formats, rates, and bit depths. Beyond that, however, they differ in a variety of ways. Some samplers use a proprietary file format, whereas others use standard WAV or AIFF files. Some use fixed directory paths, whereas others allow the samples to be easily relocated. And the list goes on.
Samplers organize their audio files similarly, mapping them across MIDI Note Number and Note On Velocity ranges, called zones (see Fig. 1). Beyond that, options proliferate and significant differences begin to appear. Some samplers automatically crossfade as you approach the boundary between one zone and another, giving you varying degrees of control over crossfade parameters. Other samplers don't support crossfading at all. Some samplers allow you to layer zones and use keyswitching, controllers, round robin, or some other technique to select among layers.
There are samplers that go beyond straightforward audio file playback, with granular processing, pitch and formant shifting, time stretching, beat slicing, complex looping, and other options. If a sample library takes advantage of specific features that are unavailable in the destination format, the converted library could either be a little lacking but sound okay or be unplayable.
Down the Path
All samplers integrate their playback engine into some sort of synthesis architecture. Most start with the standard subtractive-synthesis model, in which the playback engine is followed by a multimode filter and an amplifier, with envelope generators and LFOs for modulation (see Fig. 2). Many have a number of filter types, allow multiple filters in various series and parallel configurations, provide frequency and ring modulation at various points in the signal path, and end with a complement of effects processors.
FIG. 2: MOTU''s MachFive follows its sample-playback engine with the standard subtractive-synthesis signal path consisting of a filter, amplifier, and two ADSR envelope generators for modulation.
Modulation also often goes well beyond simple ADSR envelopes and basic LFOs. Breakpoint envelopes with dozens of stages are becoming increasingly common, as are modulation matrices for routing control signals. User-configurable performance tools such as Kontakt 2's Script Processor and GigaStudio's MIDI Rules allow real-time MIDI automation of some or all parameters.
Furthermore, the signal and modulation paths are not necessarily fixed; some samplers provide full-modulation matrices, whereas others hardwire modulation routings. Kontakt 2 has a semimodular signal path. Beyond that, different signal paths and different parameter settings may apply to different zones or zone groups within the sample map. In short, you shouldn't expect complex design schemes written by different developers to be easily interchangeable.
Now that I've made the task sound impossible, let me assure you that it is possible. Sample libraries are routinely converted to play successfully in different samplers, although there are different degrees of success (see the table “Sample Library Conversion”).
Many sample libraries don't stray far from basic sample mapping and subtractive synthesis. Almost all samplers, from the simplest to the most complex, support Note Number and Note On Velocity zones and a straightforward subtractive-synth signal path with ADSR envelopes and basic LFOs routed to the typical elements. With a library that uses only those features, all you need is a way to decode and translate the settings from one model to the other.
Not all sampler modules are created equal, and tweaking is often necessary to obtain the exact sound intended by the library's developers. Careful programming of the conversion algorithms can minimize the tweaking required, and differences in component characteristics may make a perfect match impossible — but it's not hard to get close. Unless you're going for an exact emulation of a specific instrument or sound, close is probably good enough, because you're unlikely to be using the library in different samplers side-by-side. And who's to say that for your purposes the sound you get is not better than the original? Sampler manufacturers give you all those options for a reason: to encourage you to tweak.
Some issues come up repeatedly when converting sample libraries. Once they have been identified, they are usually easy to fix. Here are some of the more common ones.
The onset of the sound may be too fast or too slow. Piano notes shouldn't fade in, and at the other extreme, you don't want clicks at the onset. The key in both cases is to adjust the amplifier envelope's attack time — slight lengthening will eliminate onset-transient clicks, and shortening will eliminate unwanted fade-ins.
Release time is another parameter that often needs tweaking. If you find notes unrealistically ringing on after you let the key up, shorten the release time. Conversely, acoustic-instrument sounds don't stop instantly; there is a natural decay as vibrations die out in the resonating body. You may need to increase the release time a bit to get natural-sounding acoustic instruments.
When the sound is too dark or too thin, the problem is probably the filter. If the signal path has a lowpass filter, its cutoff frequency, slope (indicated as dB-per-octave or number of poles), and envelope settings (especially amount and sustain level) all affect the high-frequency content of the sound, hence its darkness. For highpass filters, the same settings affect the bottom end (low-frequency content). If the sound is chirpy, look to the resonance setting with any mode of filter.
Trouble with Loops
When notes don't sustain as intended, either sustain level or looping is usually to blame. Ensure that a loop is defined and enabled, and increase the amplifier envelope's sustain level if it is not already fairly high; otherwise, notes will die out as you hold them. For ADSR envelopes, the sustain level is the S stage, but for breakpoint envelopes, you may need to do some sleuthing to deduce which stage sustains (see Fig. 3).
FIG. 3: Native Instruments'' Kontakt 2 -Flexible envelopes can have as many as 32 breakpoints. The sustain breakpoint is -indicated by the vertical orange line.
Loop settings can cause other problems, including clicking, overly obvious loop transitions, and, with extremely short loops, strange timbral and pitch artifacts. To adjust for clicks at loop boundaries, adjust the loop start- and end points in 1-sample increments. To facilitate that, you may have to zoom in to the sample level and ensure that snap-to-zero-crossing is turned off. If the sampler developer was kind enough to include loop tuning, which tunes a short loop between sample points, try that also. For more stubborn problems, most modern samplers allow you to export to an external sample editor, and I highly recommend doing so. Editors such as SoundForge, Peak, and DSP-Quattro have excellent looping tools.
If the loop transitions don't click but are still obvious, check the looping mode (forward versus forward and backward), adjust the crossfade parameters if crossfade looping is in effect, and try shifting the whole loop in either direction. To eliminate pitch and timbral artifacts, try lengthening the loop slightly. You can also edit the original sample files as necessary.
The True Effect
Modulation and effects-processor settings are most susceptible to inexact translation. Not all LFOs are created equal: you can't make a breakpoint envelope out of an ADSR, and you either have a particular effect or you don't. But assuming that you aren't trying to do the impossible, tweaking modulation settings can solve a lot of problems.
When the sound is wobbling either in pitch, volume, or pan position, look at the LFO. A subtle vibrato or tremolo can be turned into a cartoon nightmare by an errant LFO amount or rate setting. LFO-driven autopanning may or may not sound better synchronized to tempo. The wrong LFO waveform can also wreak havoc. All these parameters are subject to mistranslation and are easily fixed.
Effects processors, if they play an important role in the original sound, can be the source of significant conversion problems. Kurzweil and Ensoniq programs commonly use effects as an integral part of their sound, for example. As with modulators, the target sampler may lack the needed effects. If you end up with a sound that is too dry, add whatever effects you have that you think will work. This is an area in which you may never replicate the original sound, but you may well get something better.
You can't always get there from here; some barriers can't be hurdled. As the problems get more difficult, their solutions get more iffy. If a sample library uses a sampler feature that is not available in the destination sampler, and for which there is no viable substitute, there may not be an acceptable compromise.
Some samplers (GigaStudio 2 and SampleTank, for example) limit the number of sample references you can have for a single key range (GigaStudio 3 raises the limit to 256, which is usually enough). Others (Kontakt, HALion, EXS24, NN-XT, and MachFive, for example) allow unlimited references. You can convert from the limited to the unlimited variety, and when the limits are not exceeded, you can go in the other direction. Otherwise, you'll leave an essential part of the library on the cutting-room floor.
Hybrid hardware samplers (especially the Kurzweil K-2x series) may use audio content contained in hardware ROM. Converting sampler instruments that use that content exclusively or mixed with unrestricted content leaves only the RAM part of the content, which usually doesn't cut it.
Sample libraries that include their own virtual instruments are not convertible when the samples used in the library are encrypted. That problem arises with many Kontakt Player, Reason Refill, and HALion3 instruments, although some instruments in those formats are not protected and can be converted.
Implementation of monophonic (mono) and legato modes is another troublesome area. Mono mode limits the number of sounding notes to one, typically implementing a note-stealing priority scheme in the process. Legato mode prevents some or all envelopes (typically amplifier and filter envelopes at the least) from retriggering when a new note is played before a held note is released. Legato mode works best, and is often only allowed in mono mode. Complex sample mapping makes mono and legato modes harder to implement for samplers than for synthesizers, and many implementations are, to put it kindly, weak. But when you need it, you need it, so caveat emptor.
Many samplers allow different settings to apply to different zones or zone groups, whereas others force the same setting on all zones. That limitation can be a deal breaker. For example, Apple's EXS24 applies the same crossfade setting to all zones, which makes it impossible to convert the varied crossfading employed in many GigaStudio and Kontakt instruments. Getting the correct articulation from a layer or Velocity zone may depend on zone-specific modulation or effects settings.
The Hardware Alternative
Hardware samplers might be in the minority, but they aren't down for the count. Because they are created specifically for sample playback, hardware samplers may actually sound better and be more flexible than some software samplers. Furthermore, popular workstation keyboards such as the Yamaha Motif, Korg Triton and Oasys, Roland Fantom, and Alesis Fusion as well as drum samplers such as the Akai MPC Series and Roland MV-8000 load sampler instruments.
Chicken Systems' Translator converts most modern formats into most hardware sampler formats. Also, Translator and FMJ Software's Awave support most of the newer workstation keyboards. That enables you to convert newly released sampler instruments for use with your keyboard workstation or older hardware sampler. You can thereby avoid taking your computer to gigs and save on CPU consumption in the studio.
Remember that those structural compromises can be more radical with older samplers, especially when you are trying to cram a 100 MB instrument into a 32 MB space. Translator has some innovative schemes to deal with that, such as eliminating unnecessary fringe samples, resampling data, and truncating samples to coerce them to fit. That's often good enough, and tweaking can help.
Try It, You'll Like It
Converting between sampler formats is by nature an imperfect process and is often more art than science. As I've pointed out, that can be a good thing in providing creative options unavailable in the original library. For example, try importing a simple SoundFont into a high-end sampler — you'll be amazed where you can take it (see the sidebar “Free Samples” for sources of SoundFont and other format libraries).
For the most part, the pitfalls of translating are obvious and easy to anticipate. Don't try to convert a library that uses the esoteric bells and whistles of a sophisticated sampler into a bread-and-butter sample player. Expect to spend some time tweaking even in like-to-like conversions. Remember that not all modules (filters, envelopes, and so on) with similarly labeled controls will sound or function exactly alike. When converting to a sampler that has more features than the source, try them out.
If you have a choice of sampler formats for a library, there are a few obvious rules to follow. If you own one of the supported samplers, choose its format even if it isn't your preferred sampler. That way, you'll at least be able to compare the results. If you need to pick a foreign format, the original format of the sample library is a good choice, and the manufacturer will usually be happy to tell what that is. On the other hand, if your sampler is less complex than the original sampler and the library has been converted to a similar format, that may be a better choice because the compromises will have been made at the factory. In the best case, the library will include all supported formats; then you can try several conversions to find the best fit.
If you have the latest samplers and aren't that picky, successful conversion won't take much extra work. If you have a less capable or more esoteric sampler, you will need to take some of the steps described here. But it's worth the effort and can make your less-expensive equipment go a lot further.
Garth Hjelte is owner, programmer, and chief factotum of Chicken Systems (www.chickensys.com), maker of Translator software for the PC and Mac.
The following companies manufacture conversion software that supports most formats and samplers.
SAMPLE LIBRARY CONVERSION
This table shows the conversion status from various formats (row headings) to various samplers (column headings). N indicates native conversion by the sampler. T indicates that separate translation software is needed. N/A indicates that conversion is not available. Cells are left blank when the source format and target sampler are the same.
Click here to download a pdf of the Sample Library Conversion Table.
You can find free sample libraries of almost any kind of content on the Web. Here are some useful sources.
Lots of great SoundFonts covering a variety of General MIDI sounds, and other synthy, ethnic, and percussive sounds.
Freebies are posted here monthly.
This site features sounds in EXS24 format.
A small collection of Worra's sounds, but offerings are changed regularly.
Original ethnic-flavored sounds in EXS24 format and precisely trimmed WAV files. This site also has classic, synthesized drum samples.
Exotic percussion and guitar instruments in Giga format.
A variety of keyboards, guitars, and percussion instruments in Giga format. (You are required to create an account before accessing the free area.)
Extensive collection of synthy-sounding EXS24 instruments. This is a German-language site.
Dirty, lo-fi drum hits, loops, pads, and instruments for EXS24. The drum hits are easy to convert to other sampler formats.
Large, well-organized collection of SoundFonts covering all instrument categories.
Synth-based SoundFonts with emphasis on pads, strings, and percussion.
Casio VZ-1 and Yamaha RM50 samples reformatted as EXS24 instruments.
Samples created from Reactor Ensembles and converted to EXS24 format.
Two great-sounding drum kits converted to a variety of formats.