One of the software categories that has benefited most from recent advances in computing power is software synthesizers. Today, turning your computer

One of the software categories that has benefited most from recent advances in computing power is software synthesizers. Today, turning your computer into a full-blown synthesis engine is nearly a trivial matter, and even systems that are less than state-of-the-art can usually provide valuable new resources to the desktop composer.

This article will examine a wide range of programs that use your computer's CPU as the sole sound-generating hardware, often with all the adjustable parameters that you'll find in today's outboard synthesis gear. I'll look at the most recent versions of some older programs and discuss several of the hottest new releases on the market.

Because so many programs are available, I'll only cover a few in each of the main categories. No doubt you'll find many other excellent programs, especially shareware, by surfing around the Net. Also, my focus will be on Macintosh and Windows applications, but you can certainly find high-quality options for other platforms such as Linux and BeOS. And I'll discuss only programs that run free of any dedicated hardware, which means that plug-ins written for TDM, for example, won't be included.

THOSE WERE THE DAYSThe concept of native signal processing, in which the computer's CPU provides all the computing power that a real-time multimedia program needs, is not new. In fact, it was originally brought to market by Intel, Motorola, and others in the early to mid-'90s-alas, somewhat before its time. Only in the past few years, with the advent of vastly more powerful processors, has it been possible to realize the potential of native signal processing, and all of these programs operate under that basic process.

Of course, when Max Mathews at Bell Telephone Labs first developed the means for computers to make sound in the late 1950s, the method of specifying values for sound parameters was much different from what it is today. Mathews and his colleagues used punch cards to enter their scores, as well as the routines that the computer used to generate sound. That method remained current for several decades, until terminals on which data could be typed and sent into the computer became available. (See the sidebar "Other Avenues" for a description of two modern programs that stem from Mathews's work.)

What a long way we've come. The soft synths covered here offer a host of real-time controls, including onscreen sliders, knobs, and faders. They also can be controlled from external MIDI keyboards, joysticks, and the like, allowing you to "play" them live. And in most cases, you can control them from a sequencer running on the same computer.

YOU PAYS YOUR MONEYWhat are the advantages of using a soft synth? Sound quality comes to mind; if your audio hardware supports digital I/O and has multichannel outputs, you can look forward to excellent audio quality with considerable routing flexibility. Another advantage is independent effects processing on each MIDI channel, a capability you'll get from several of the multitimbral synths. With so many basic design components available, you can also expect a high degree of sound-design flexibility; and of course, adding new features by way of software downloads is a cinch.

But there's no question that dedicated synthesis hardware also has its advantages. These include reliable and consistent performance (because it's not dependent upon a host CPU), portability (rack it and take it on the road), low or nonexistent latency, and ease of real-time tweaking (just how many parameters can you change at once with a mouse?). Nevertheless, a synthesizer running completely in software can be the best solution for many musicians, and it will certainly be a tremendous asset even if it's not your only means of creating sound.

By the way, I won't be making any absolute judgments about performance here, because it varies greatly depending on many factors (see the sidebar "Performance Practices"). There are simply too many variables to establish any relevant benchmarks.

However, on the test beds that I used-a Pentium II/400 MHz with 128 MB of RAM and a Mac G3/266 MHz with 256 MB of RAM-I got completely satisfactory performance from all these programs when using them as my only sound source. (Using a soft synth while playing audio tracks in a sequencer or attempting to sync incoming MIDI data with events in a sequencer track being routed to a soft synth can be more problematic.) I'm confident that a soft synth can offer valuable sonic resources on any modern machine; but, of course, the more horsepower you can put at your synth's disposal, the better your results will be.

SYNTH CONSTRUCTION SETSDozens of software synthesizers are available today, and for the most part they fall into several basic categories. The first group that I'll examine consists of programs that are, in effect, synth construction kits. They offer building blocks for creating your own sound modules, and one gives you tools not yet available in any of the most modern hardware synthesizers. The synths in this group run as stand-alone applications, but some also work as plug-ins, and most are intended to be operated live using MIDI note and controller data or to be triggered by a sequencer.

These programs share other attributes. For example, most, if not all, can write their output directly to disk, giving you a 16-bit pure (or better) signal path straight to your CD burner. (Many programs in other categories have this same capability.) The majority of the synth construction kits have some type of built-in sequencer for triggering sounds and accepting MIDI input. And though you'll find significant differences among the interfaces that they offer, each has some form of modular, "patchable" design layer in which you connect various components to create your final, unified design.

Here, then, is a look at several programs that provide a modular approach to sound programming with real-time MIDI control.

Tassman 1.2 (Win; $395). One of the most interesting new modular synths is Applied-Acoustics' Tassman. Other synths have used physical modeling in the past, most notably Seer Systems' Reality, but Tassman's catalog of sound-producing options is unique. Here you'll find bows, bars, beams, and many other objects with which to build virtual instruments (see Fig. 1). These objects open up a vast realm of sonic possibilities that you won't find in any of the other programs reviewed here.

Tassman splits its functions into two independent applications: Builder, with which you design and test your synths; and Player, which you use to play them. This is a bit unusual and somewhat unwieldy; although you can load Player from within Builder, it would be easier to have a single screen from which you could load any of the dozens of presets, play them, and modify their designs. Instead, you typically access synths directly from an Explorer window, then choose whether to run them in Player or load them into Builder for modification.

Tassman's sound-design tools are extensive but not as numerous as those in some of the other programs. On the other hand, Tassman's unique physical-modeling modules really give the software its identity. Interested in the sound of a marimba being bowed or a flute being struck by a mallet? In Tassman's world of virtual instruments, these and many other anomalies are possible. Moreover, the parameters that the modules use, such as pick stiffness and mallet strength (force), are intuitive and easy to control.

Applied-Acoustics gets high marks for its documentation, which includes a well-paced tutorial and a good explanation, along with suggested usage tips, for each of its modules. The coding also seems very efficient; I was able to achieve high levels of polyphony with many of the presets. If you're looking for some new sonic dimensions to add to your work, Tassman could be just right. It's one of the most exciting new programs that I've seen in a while.

Reaktor 2.3 (Mac/Win; $499). Native Instruments' software is noted for its power and flexibility. The company's latest offering is Reaktor 2.3, which combines the synth application formerly known as Generator with the company's sampling workhorse, Transformator. (these products are no longer available separately.) This potent combo provides a massive array of sound-generating and sound-processing modules that you can freely connect to build your synth designs. A huge number of presets are also included, created by both the company and Reaktor users.

Reaktor's synthesis options run very deep. To access any of the dozens of modules, you open a new Ensemble and insert the components that you need. Modules come in various levels of complexity: some are macros that perform several functions; others are more basic and can handle only a single task. The toolkit includes functions for many different mathematical operations and for more familiar sound-generating processes.

Once you have completed an Ensemble, you switch to the Panel display in order to play it and adjust its parameters (see Fig. 2). In the Panel, you can take snapshots, which capture the status of all the parameter values at a given moment. Snapshots can be accessed from a drop-down menu and via Program Change messages.

Reaktor's sampling features are well integrated with its synthesis functions and are a huge resource in their own right. The combination of a vast array of sound-generating modules and numerous sound-transformation features makes Reaktor an excellent choice for a wide range of tasks. Its ability to function as a VST Instrument is simply icing on the cake. And if you get the free Premium Library (an excellent collection of presets), you'll be ready to run dozens of patches right out of the box.

Buzz 1.2 (Win; free). Oskari Tammelin's Buzz is a modular synth builder, but it looks and feels quite different from most of the programs in this roundup. It appears to enjoy a wide following; many of the bundled presets were created by users, and the Web site is an active spot for exchanging information about the software and related topics.

There are two main categories of synth-design modules: Generators and Effects. Buzz calls these modules machines, and you build complex synth networks by wiring the machines together. To control the machines, you create Patterns that remain fixed or vary over time. Many (but not all) machines can also employ MIDI input either alone or as a "live accompaniment" to the control data contained in the Patterns.

Patterns are created in the Pattern Editor and then scheduled in the Sequence Editor. The Pattern Editor is, in effect, a sequencer that can be used to control the time-varying evolution of any machine parameter. This means that you can animate the wet/dry mix of a reverb, the depth of a flanger, or simply the frequency of an oscillator. Tools to help you tweak your Patterns are provided; for example, the Interpolate function creates a series of values from a given start point to an end point. The Randomize feature alters the values in a range. Many of the values that you supply for your Patterns must be hexadecimal, so get out your calculator if you're not comfortable with base 16.

You can insert any number of Patterns into the Sequence Editor and then set the Editor to one-shot playback or make it loop over a selected range. You can adjust the tempo as a machine plays and view a display that continuously updates the value of any chosen parameter. A CPU monitor allows you to see how hard your computer is working during playback, and you can also watch a real-time frequency analysis of the audio being generated.

Buzz seems to play well with others; it can load VST plug-ins as effects, and there's even a browser plug-in available that lets visitors to your Web site hear your Buzz creations. You can also load Buzz files into Nullsoft's Winamp using a free plug-in. Buzz allows you to build your own basic synth components, but you'll need some programming chops and a Visual C++ 5.0 compiler for the task. However, the overall flexibility of this program could make it a major player in your audio environment.

Retro AS-1 2.0 (Mac/Win; $259). Bitheadz has recently released a major upgrade to its Retro AS-1 software, and this is one of the most potent tweaks in memory. There are numerous enhancements; for example, version 1.0 offered a maximum of 32 voices, but the new 2.0 bumps that up to 64. The company has also added new filter types, new insert effects, and significant enhancements to the interface. You can even route external audio into the program and use it as a modulator in the new Ring Modulator section. As a bonus, Retro AS-1 is fully ReWire-compatible. (A separate program version-Retro AS-1 VST-can be used as a VST Instrument on the PC.)

In Retro AS-1's Configuration window, you view and add or remove modules in your synth design (see Fig. 3). Clicking on any module brings up the Main window, where you can set values for any of the module's parameters. Mod sources and effects (up to two insert and two global per patch) are configured in their own respective windows. In the final screen, Global, you choose from various voice-activation, portamento, and MIDI-related options. (It takes a bit of window jumping, as you might imagine.)

Each of Retro AS-1's three stereo oscillators can use one of ten waveforms or three types of noise. For those who really like to drive their sounds hard, the software offers numerous modulation and sync sources. Also available is a Random tuning parameter, which you can use as a modulation destination if you want your sounds to have the classic analog "detuned" quality.

The VCF section contains an assortment of 1-, 2-, and 4-pole filters-with and without resonance-along with the traditional state-variable group (lowpass, highpass, bandpass, and the like) and threshold, comb, and slope. The filters can be configured to run in series or in parallel. An Overdrive parameter is offered for those whose tastes run to the truly gritty, and each filter's cutoff frequency can be modulated, giving you even more possibilities for nonlinear distortion. Two 5-stage envelopes and a dozen internal effects round out the basic toolkit.

Retro AS-1 wins the prize for most presets with more than 1,300, grouped in categories such as FX, Techno, Strings, and Drums. They range in sound from the incredibly fat and juicy "Alaska," to the punchy "Rezo Kizz Bass," to the delicate "Moonflower." I only wish that I could audition them directly from the File dialog box so that I wouldn't have to load them to try them out. In any event, the software clearly provides a huge range of sonic possibilities. It's also well acquainted with all the standard Mac audio drivers, which means it should integrate nicely into any studio environment. And if you happen to have a G4, you'll find that Retro AS-1 is highly optimized for performance on that system.

SynC Modular 1.7 (Win; $49). Dr. Sync's SynC Modular has an interface that resembles Reaktor's, and it uses similar terms for many of its design components. The program groups its modules into several categories, all of which can be accessed from a drop-down menu by clicking the right mouse button. In the menu you'll find Basic modules, which are SynC Modular's most fundamental building blocks; Standard modules, which are the next level of architecture and typically contain many Basic modules as their source; and Library modules, which include macros contributed by users in addition to those provided by the manufacturer. You can create your own macros by loading a Macro module and then building whatever structure you want it to have.

Each of the modules has a host of parameters that you can access and edit by right-clicking on the module's icon. In addition to its sound-making components, you can customize many other aspects of a module-for example, whether it appears in the Panel display, what caption appears on its icon, and, in some cases, what bitmap that icon uses.

A number of shortcuts make revising your designs easy. For example, when you're working in the performance-level Panel display, you can click on any icon and have SynC Modular jump directly to the basic component that the icon is controlling, even if it's many layers deep in the synth's structure. The program's ability to render MIDI files is especially useful, and its support for DirectSound ensures that you'll get the most mileage out of your audio hardware. It can also be used as a VST Instrument.

SynC Modular's documentation is available only in a Help file, but it's surprisingly thorough. The documentation includes numerous optimization tips, which cover both specific design issues and global computing considerations. The program's design options are very deep and allow you to work at an incredibly low level of architecture (though you may need to sharpen your DSP chops if you really want to get your hands dirty). Moreover, the program performed exceptionally well on my test platform-equal to, if not better than, most of the others in this roundup. At $49, SynC Modular may be the best bargain in the bunch.

VAZ Modular 2.1 (Win; download, $282.27; CD-ROM, $338.69). Martin Fay's VAZ Modular hides much of its enormous power behind a simple interface. Rather than use virtual patch cords like most of the programs in this group, VAZ employs drop-down menus for making connections among the various modules in your design (see Fig. 4). It also has a nifty feature that can "roll up" an individual module so only the title appears. This allows you to create a clean, lean look for your synths and to keep the interface well organized.

The program offers two playback methods: real-time MIDI control and a built-in sequencer. VAZ's sequencer is a very deep feature that has many layers of operation. It offers 16 patterns, each of which can contain up to 16 steps. Each step of a pattern has a Pitch, Rest (mute), Slide (portamento), Accent, and Double (for doubling the step's length) control. You can gang all of the controls-for instance, if you want to transpose the pitch of your entire pattern-and adjust the pattern's tempo (from 1 to 255 bpm). But that's just for starters.

The next level of operation controls pattern playback. You can play all patterns in order or set a range (for example, to play back patterns 2 through 6). You can also play patterns in reverse order or use the Song Editor to create a sequence of up to 255 patterns in any order with a unique transposition offset for their pitch. Finally, a pattern randomizer feature allows you to generate patterns randomly and to constrain the options that the randomizer uses. You can limit the range of notes chosen, the number of steps used in each random pattern, whether rests are allowed, and more. If that's still not enough control, you can use a full range of real-time MIDI triggers to determine which patterns will play.

VAZ offers some very powerful sound-design components, including a full complement of audio-generating and audio-processing tools. Among the more unusual are the granular oscillator, which generates a stream of grains at either a fixed or variable frequency; the vowel filter, which morphs between different vocal sounds; and the envelope follower, which converts an audio signal into a control-rate signal. Although its internal effects are fairly limited, VAZ is one of the few synths in this roundup that support both DirectX and VST plug-ins. Overall, it's an elegant program that should be high on every musician's list of considerations.

BEAT THE BOXThe stand-alone applications in the next group don't let you design your own synths from scratch; instead, they offer screens full of tweakable parameters that control the generation of sound. Here you'll find familiar sound generators and processors such as LFOs, VCOs, and filters for tweaking your sounds; and for the most part, MIDI control is extensively available. All of these programs model analog synths in one form or another, and several use the "beatbox" model popularized by Propellerheads' seminal ReBirth program. Arturia's Storm even lets you load several different "analog" tone modules at once and synchronize them to a global clock.

Vibra9000 (Mac; Studio9000 bundle, $595). Vibra9000 is part of Koblo's Studio9000 bundle and is the easy winner of the most-colorful-software award. Its bright green interface features a single resizable screen on which are knobs for controlling all the program's parameters (see Fig. 5). You can easily adjust the knobs with the mouse (though you can also use the preassigned MIDI controller for any of them), and a large display indicates the exact value of every parameter you're tweaking. Drop-down menus provide access to other program options.

Vibra9000 groups its functions according to a traditional analog synth design. You can assign any number of waveforms independently to its two VCOs-a knob produces a continuously variable change in the waveform's spectrum as you slowly move it from, say, triangle to sine. Two separate LFOs offer a choice of six waveshapes, and the three ADSR envelopes can be applied to numerous sound parameters. Individual envelope segments can be finely adjusted using 0.001-step increments in a range of 0 to 1, but you can't type in an exact value.

The filter section features 2-, 4-, and 8-pole options along with double, quad, and notch types for added variety. It also offers the unusually named SawComb and SqrComb, which provide multiple resonant peaks that resemble the waveforms suggested by their names. The Distortion parameter is useful for creating sounds that fall into the "nasty and raunchy" category, and the Keytrack setting allows you to determine how closely a filter's cutoff frequency maps to an incoming MIDI note value.

Whipping through the hundreds of presets is made nearly painless with the Page Up/Page Down shortcut (which loads successive patches), and by using the Trigger and Hold controls in the Global screen you can trigger a sound and have it play indefinitely. The Arpeggiator offers several modes for cycling through the preset note patterns.

Vibra9000 provides drivers for nearly every Mac audio protocol there is, but many of the drivers are still in beta. (Check the company's Web site to see if your hardware is supported, so you can avoid using Sound Manager if at all possible.) You can also use Vibra9000 as a VST Instrument, and it's especially happy when running with DirectConnect (for use with Digidesign hardware). And though Vibra9000 is strictly monophonic, its high-quality sound generators and modifiers-coupled with extensive routing options-make it an excellent tool for creating a wide range of analog sounds. It is also a sheer pleasure to tweak. Use it along with the Stella9000 sampling and Gamma9000 drum machine modules (also included in the Studio9000 bundle), and you'll have a potent pack of sound sources, all living under the same roof.

Reality 1.56 (Win; $379). Seer Systems' Reality was the first professional software synthesizer to hit the market, and though it's showing signs of age, it remains one of the most potent players available. The program's strength lies in the high quality of its physical-modeling technology, which can produce an extensive range of accurate acoustic simulations. You'll find excellent-sounding wind, percussion, and string models, including very realistic flutes, clarinets, marimbas, chimes, and plucked strings. Reality's alter ego adds an entirely different set of features to the program: it can also load, manipulate, and play back samples in both WAV and SoundFont formats.

In addition to the physical models, Reality's other main sound source is its group of four oscillators, each of which can be assigned one of six waveforms or three types of noise. The Topology parameter allows you to configure the oscillators in an additive network or in several types of FM arrangements. Each oscillator includes a gain and coarse-tuning parameter-both of which can be modulated by a large number of internal sources or MIDI controller data-a random pitch offset, and an adjustable Velocity-response parameter.

Eight filter types are next in line, followed by four LFOs and four multisegment envelopes. No graphic interface is provided with which to draw envelope shapes; instead you pick one of the default shapes and modify the time or level of the segments using numeric values. Like most parameters in the program, the LFOs and envelopes can have up to two modulators of their own, and you can easily configure the modulation routings using the pull-down windows, where all the options are displayed.

Reality has also overcome one of the major problems in using a stand-alone software synthesizer with a digital audio sequencer, namely, that many software synthesizers take control of your sound card and don't allow the playback of audio from any other source. Using the included Reality Wave Driver, you can route audio from your sequencer directly into Reality, mix it with Reality's own audio output, and send the combined signal to your sound card.

A recent update to Reality added support for a number of new sound cards, enhanced support for SoundFonts, and solved some user-reported installation problems. It also gives you the option of exporting an entire bank of patch names in the instrument-definition formats of many popular sequencers. Though its interface could use a major overhaul, this robust program has much to like, and the many synthesis options, not to mention the excellent physical models, make Reality suitable for a huge range of music projects.

Storm 1.1 (Win; $225). Arturia's Storm uses an interface model that is different from the programs covered so far. It offers a "rack" into which you load multiple synth and effects modules that are controlled from its elaborate Composition screen (see Fig. 6). Analog synths and drum modules are the story here; Storm also has several sample players and a "turntable" on which you can "scratch" samples via MIDI control. Hard disk recording of the program's output is managed very nicely and is well integrated with its other features.

A Storm rack can contain up to four modules, each of which can be routed through up to three effects units. (The effects are Delay, Chorus, Distortion, Sequence Filter, and Flanger.) All the synth modules include pattern-based sequencers for triggering notes and provide various controls for adjusting their sound parameters. For example, Arsenic is a bass-line synthesizer that displays a 3-octave keyboard on which you create note patterns. The keyboard can be transposed to provide a 12-octave total range, and each of the 16 notes in a pattern can have one of four Velocity values. Arsenic's sound source is a single oscillator (square, triangle, or pulse) that is hardwired to a filter. The synth includes controls for attack, decay, filter cutoff, resonance, and cutoff modulation.

You mix the output of your rack's various synths in the Composition window, where you'll also find the Editing Bar and several other composition tools. The Editing Bar is a visual representation of your entire composition, with small squares representing individual bars of music. Here you make changes to global parameters (such as tempo or volume) or record changes to the parameters of individual synths. A clever feature called Kepler is used to transpose your patterns on the fly or to create a sequence of key changes. (Storm can modify the pitch and tempo of audio in its sampling modules as well as the output of its synths.)

You can record the output of one or more sound modules using Storm's Recorder feature and then immediately use the recording in your compositions. And the entire operation can be done while your piece continues to play-an especially nice feature. Storm's sample-playback options are beyond the scope of this article; suffice it to say that there's a great deal of visual feedback, which makes working with multiple samples quick and easy, even in real time.

Storm is clearly best suited for dance, techno, or any genre that relies on looping patterns. But the real-time pitch-shifting and time-stretching that you can perform on samples adds another aspect to its personality, one that can lead to many other realms of music making. Storm can be used as a VST Instrument, and it's one of the few programs in this article that includes a paper manual. Though its initial release has some rough edges, Storm is a very capable performer that can bring you hours of enjoyment.

SimSynth 2.6 (Win; $35). Image Line's SimSynth is a vintage-analog-synth emulator with a slew of modern features. It has a number of useful compositional tools as well as some graphic elements that make designing patches quick and intuitive.

The VCO section includes three oscillators, two of which can serve as modulators in an FM configuration (see Fig. 7). There are five default wave shapes (saw, square, pulse, noise, and sine), but an Adjust knob allows you to create crossfades among the spectra of the first three. If you select sine, the Harmonic Synthesizer becomes accessible; here you can build composite waveforms using any of 16 harmonic partials. Though you can't type in exact values for the program's various parameters, a clear display shows the current value of the parameter you're adjusting.

Every parameter in SimSynth can be mapped to a MIDI controller; you just right-click on the knob and assign it a CC number. (you can also change the range of values that the knob transmits.) For those of you who are graphically inclined, user-created envelopes are available to control pitch, filter cutoff, or amplitude. (all three can be used simultaneously.) By zooming the display in or out, you can make very accurate adjustments to individual envelope segments. SimSynth also provides a built-in sequencer-complete with a metronome-for recording and playing back MIDI data, and a 2-octave keyboard enables you to trigger notes and create patterns.

All output from SimSynth can be captured to disk, including the sound triggered by any MIDI data that you've recorded. The many excellent presets, which begin to play when they're first loaded, are a great resource for studying good patch design. (You can get some really sassy bass sounds if you set the filter cutoff to around 10 percent and experiment with the Adjust knob while using a pulse wave.) To help you build your own presets, the program has a Preset Quick Starter feature that includes suggested settings for various types of sounds, such as Electronic and Brassy-very smart. In fact, that's a good overall description of this well-designed application.

Probe 1.5 (Win; $199). A fairly new arrival in the beatbox world is Synoptic's Probe. This capable, 16-voice synthesizer offers up its three oscillators, two envelopes, and two filters in a single main work area. However, lurking beneath the surface are some unique tools that raise the level of functionality a good bit. For example, clicking on VCO I, VCO II, or either of the LFOs opens the Additive Synthesis window, which contains sliders for the amplitudes of 20 harmonic partials. By adjusting the sliders, you can create and then save any number of complex waveforms for use in your designs. You can also view the composite waveform as you adjust the levels of each partial.

Another of Probe's unique features is its ability to analyze the first 20 harmonic partials of a sample file on your drive and load the resulting spectrum into an oscillator's wavetable. Once the spectrum is analyzed, you can alter the amplitude of individual harmonics, which gives you a very simple analysis/resynthesis module. You can also morph between two waveforms, and you can even control the morphing rate with an envelope or LFO.

And speaking of control, you can route an audio signal into Probe and use it in various ways. You can employ the audio as a sound source in place of a VCO, or you can extract the amplitude envelope of the audio input (your voice, for example) and use it to control a parameter such as the VCF cutoff. That way, the louder you spoke, the higher the filter's cutoff frequency would be.

Probe's sequencer is very powerful and lets you place audio files or synth patches onto any of its 17 tracks. Up to 16 patterns can play back as many as 24 steps each, and any number of synth parameters can be sequenced over time. It can also import a MIDI file and extract information from one or more tracks for use in your own patterns.

Program Change maps can be easily set up to recall any of Probe's presets, and a large number of options are available to configure the program for use with other MIDI gear. Its high level of customization and numerous high-quality audio components make it an excellent choice for your analog needs.

ALTERNATE MODESMost of the programs in the next category accept MIDI data for the control of various sonic parameters but are not primarily intended for use with a keyboard controller. Instead, knobs, sliders, and other interface elements are available for triggering playback, and you'll also find the occasional automated control to relieve you of some of the real-time duties.

These programs all generate sound in real time using various synthesis methods, some of which are far less common than those used in the programs I've covered so far. The big winner here is granular synthesis, which is at the heart of several members of this group. Though at first you might suspect this category of being on the fringe of the musical landscape, you'll be surprised by the range of sonic options that the following programs provide.

GranuLab 8 (Win; $20). Although granular synthesis is one of the most efficient ways to generate unique and colorful sonic landscapes, it's a method that hasn't yet appeared in major hardware devices. (Symbolic Sound's Kyma System is the one exception.) However, granular-synthesis software is another story. Rasmus Ekman's GranuLab 8 is one of several powerful granular synthesizers, and, like some of the others in this group, it allows you to synthetically generate thousands of small sonic grains or to slice and dice existing sound files into minute fragments.

GranuLab 8's single screen consists of numerous sliders that control the various parameters of each grain. In addition to sliders that set the base value for pitch, grain density, amplitude, stereo position, and glissando rate, there are separate sliders (one for each base slider) that add an amount of randomness to each parameter's base value. (When using samples, you can also assign a value to the sample start position and add a random offset to that point for each file.)

You can run up to eight different sets of controls, each producing a separate grain stream, and you can mute or solo individual streams. Depending on the density of each stream, that may be more than many computers can manage. You can also save 20 snapshots of your settings at any given point and instantly toggle between settings, or you can set a parameter change time to determine how long the transition time from one snapshot to the next will be. (GranuLab 8 morphs between the two settings over the time that you indicate.)

Additional real-time control is available in the Patch Gesture window. This window, which functions like the Vector control in some Korg devices and the Blue Window in Arboretum's Hyperprism, lets you use your mouse to "morph" among four different snapshots. As you move the mouse to extreme corners of the window, the snapshot you've assigned to that corner takes precedence. Finally, you can assign a Program Change to recall snapshots, or you can simply use MIDI controllers to manipulate most parameters.

A freeware version of GranuLab 8 with only one grain stream is available at the developer's Web site, but if you can see a world in a grain of sound, pick up the registered version and experience the entire universe.

CrusherX-Live 1.4 (Win; $39). Another contender in the granular group is Joerg Stelkens's crusherX-Live. This awesome little application has a host of features for granulating samples of real-time audio input or generating grain clouds from scratch. You can even mix multiple sources in the same session. The program's parameters can be controlled using several types of MIDI messages, and it's well suited for use with alternate controllers such as a joystick. Though the interface won't win any prizes for elegance, the sliders and buttons are easy to manipulate, and the program's overall design is very intuitive.

At the heart of crusherX-Live is a set of oscillators that are summed and sent to the Crusher panel (see Fig. 8). The panel uses the oscillators for input and offers numerous controls for their manipulation. Like many implementations of granular synthesis, crusherX-Live allows you to control grain frequency, amplitude, timbre, length, density, and time between grains. Random offsets can be added to any parameter using several types of random functions; and you can generate multiple grain streams simultaneously. Several windows provide a graphic display of the state of the various parameters.

CrusherX-Live excels at real-time performance and is a powerful instrument once you master its intricacies. For example, the Loadlist offers the functionality of a playlist; you can use it to load numerous preset parameter settings and switch among them-manually or using MIDI events-as the program plays back. With a set of buttons at the bottom of the screen you can randomly assign a new set of values for one or more parameters, and the unlimited Undo/Redo (which can be enabled via MIDI notes or with a button on a joystick) lets you return to any point in your session.

The program's ability to interact with a force-feedback device (whereby the value of different parameters causes you to feel greater or lesser pressure at the controlling device) is yet another unique feature that encourages live experimentation. CrusherX-Live can produce an enormous range of interesting textures, and one of them might be just right for your next gig.

Chaosynth 1.0 (Mac/Win; $40). Nyr Sound's Chaosynth offers one of the most unusual means of sound generation that you're likely to find. The program combines granular synthesis with an algorithmic process known as cellular automata and allows the user to "perform" many of the synthesis parameters in real time. The mouse is the primary control tool, but MIDI controllers and even Note On messages can be used to alter many parameters as the synth plays.

Cellular automata is a process that employs a set of rules to determine the way in which a series of values evolves over time. As implemented in Chaosynth, the process makes determinations about what notes will be heard at a given moment and governs the transition from one set of parameter values to the next. The result can be anything from massively dense grain clouds to subtly evolving, almost organic textures that swirl throughout the spectrum.

User interaction occurs in many ways. For example, you can drag the mouse in the Frequency panel to change the number and range of frequencies that are included in the grain stream (see Fig. 9). Or you can draw curves in the Oscillator panel to alter the amplitudes of the various frequencies. Grain size is controlled in its own dedicated window, and among the many options are controls to alter the behavior of the automata process and to apply reverb, filters, and envelopes.

One of my favorite features is the ability to convert grain events into MIDI notes, which means you could have one layer of sound that uses the basic waveforms in Chaosynth while another layer plays a MIDI-orchestrated version of the same data. The possibilities are endless.

Chaosynth includes an excellent tutorial on the synthesis and algorithmic methods it uses and provides numerous interesting presets to get you going. If you're looking to add an entirely new dimension to your audio toolkit, Chaosynth is definitely worth checking out.

WaveWarp 1.2 (Win; $199). WaveWarp, from Sounds Logical, is a massive construction kit for building synth networks and signal-processing routines; and though it doesn't yet accept MIDI input, it offers numerous ways to control the generation of sound in real time. For example, you can easily build complex additive-synthesis networks that have controls for each partial's amplitude or create subtractive synthesizers using any of the program's various high-quality filters. You can also make "self-running" FM designs that generate hours of slowly evolving ambient textures.

WaveWarp's modules are accessed through tabbed folders that appear at the top of the main screen. You build a synth network by dragging modules onto the Drawing Board, where you make the connections and set the parameters for each component in your design. Clicking on any component brings up a window where its parameters can be controlled as the sound plays.

The list of sound-generating modules is exhaustive. In the Signal Generating category, you'll find sweep generators; sine, square, and triangle wave generators; chaotically controlled signals; and more. Reverbs, pitch-shifters, and spectral transformers are also provided in large quantity, and delays, choruses, and flangers are just a few of the many additional tools. Don't expect any automated patch-generation options here, though; WaveWarp is a serious sound programmer's toolbox, and, as such, it demands a fair amount of knowledge on the subject.

Using WaveWarp as a stand-alone synth would keep you busy for some time, but the new 2.0 version, which should be shipping by the time you read this, allows you to use your creations as DirectX plug-ins. Though it isn't currently suitable for live performance using a MIDI controller, it has an enormous range of powerful sound-producing modules that you can combine in an endless number of configurations. If you're a hacker at heart, WaveWarp should be in your workshop.

PLUG-IN SYNTHSRunning a software synthesizer from directly within a host application is now quite popular; Steinberg deserves the lion's share of the credit for making its VST Instrument format the method of choice for nearly all software of this type. Dozens of VST-format synths run as plug-ins, and using them can be as simple as assigning the output of a sequencer track to the VST Instrument and pressing the Play button. (Different hosts-for example Cubase VST and Logic Audio-use different means of employing VST Instruments.)

VST Instruments offer several special advantages. For example, you don't need to use virtual MIDI drivers or IAC connections because the Instruments receive their control data directly from sequencer tracks. You can also automate all of your synth patches' parameters using MIDI controllers, and you can usually save complex synth designs right along with your sequencer files. Furthermore, you don't have to worry about which internal effects a VST Instrument offers, because you can route its audio output through any VST or DirectX effects plug-in on your system.

You should make sure that you have current, stable audio drivers installed to get the best performance from a VST Instrument; and remember that you can't tweak the latency of a VST Instrument separately from that of the other audio sources you're using in your digital audio sequencer. As stated earlier, don't expect to stay in sync with the music while performing a VST Instrument live from an external keyboard and while multiple tracks of audio are playing back (much less multiple effects). Most systems simply can't offer that level of performance yet.

With the exception of Emagic's ES1, all of the synths in the next group are intended for use as VST Instruments. Though VST Instruments come in several varieties, most model classic analog synthesizers. (An excellent resource is Ben Turl's K-v-R VST Instrument Banks site at www.k-v-r.freeserve.co.uk. Here you'll find the occasional virtual bass or drum machine as well as early announcements of new synth plug-ins.) Keep in mind that many of the synths I've covered so far, such as Native Instruments' Reaktor, can also be used as VST Instruments. That gives you even more options to use in this format.

If you're a PC user and you don't have a VST 2.0 compatible application, you'll be pleased that VST Adapter from FXpansion (www.fxpansion.com) lets you load a VST Instrument into many DirectX-compliant programs. (Problems have been detected with certain DirectX hosts. Ask your software manufacturer about any known incompatibilities.) Regardless of the host program, you'll find that these plug-in synths offer considerable editing flexibility and integrate extremely well with your sequencer environment.

Model-E (Mac/Win; $199). Everybody knows that analog modeling has been a hot ticket for the past few years, but no one could have suspected that so many legacy hardware devices would be modeled directly in software. Among the plug-ins that model vintage hardware is Steinberg's Model-E, which has the look and feel of a Minimoog Model D (see Fig. 10). No doubt you'll find that this virtual version stays in tune a lot better than the original!

Model-E offers enhancements that include expanded polyphony, complete MIDI control, and 16-part multitimbral capability. In place of the original Moog's keyboard, you'll find a MIDI setup area where you pick programs, banks, and MIDI channels, and assign the amount of polyphony that you want to dedicate to the synth. (The maximum is 64 voices.) You can send notes directly from an external keyboard, but most likely you'll trigger the synth from a track in your host sequencer, just as you would with any of the VST Instruments discussed here.

Each of Model-E's three oscillators offers independent level control and a choice of six waveforms with a frequency range that spans 6 octaves. Oscillator 3 can be used as a mod source, such as an LFO. A noise generator provides an additional sound source and is especially useful for percussive timbres.

The VCF is switchable between 2- and 4-pole and can track the pitch or Velocity of incoming MIDI notes. The filter also has a dedicated envelope generator and can be modulated by oscillator 3 or from your controller's mod wheel. Glide and Tuning settings are found in the Controller section, where you also enable the different modulation options.

Certain trade-offs are always required when modeling hardware devices in software, but I still feel that the control of Model-E's rotary knobs might be improved. Unless you click on the very tip of a knob when you adjust it, the knob jumps well beyond its current setting, making it difficult to fine-tune a parameter. One good solution is to use a physical MIDI controller to make adjustments, which also allows you to tweak several controls at once.

Regardless of the minor problems, Model-E is a massive achievement in modeling technology and brings the sound of one of the most popular synths of all time back to the studio. It should conjure up fond memories for quite a few people.

Pro-Five (Mac/Win; $199). Pro-Five from Steinberg/Native Instruments is a virtual model of Sequential Circuits' Prophet-5; and like Model-E, it adds significant new features to the original (see Fig. 11). Pro-Five tops off at 32 voices, which should be well beyond the ability of most host systems. (If you need additional voices, you can load multiple instances of the instrument.) For ultrafat sounds, you can put the device into Unison mode, in which all available voices play a slightly detuned version of the same note.

Like the Prophet-5, Pro-Five provides two oscillators per voice, each of which can be tuned over a 4-octave range in semitone increments. Oscillator A offers sawtooth, pulse, or a combination of the two; oscillator B adds a triangle option. (If you select the pulse wave, you can vary its duty cycle from 1 to 99 percent.) Oscillator A can be hard-synched to oscillator B, which also doubles as an LFO. As with Model-E, you can add a noise generator to the mix.

Triggering notes on Pro-Five is especially easy, as the plug-in sports a built-in retractable keyboard complete with pitch-bend and mod wheels. Though the knobs are fairly small, moving them with the mouse is easy; just grab a knob and drag the mouse up or down. (a rotary-movement mode is also available). Numerous modulation sources-mostly controllable from the Poly-Mod, LFO, and Wheel-Mod sections-provide the full range of sound-design options that polysynth users have long enjoyed.

Pro-Five ships with 512 sounds that are organized by Program, Bank, and File. But unlike Model-E, Pro-Five doesn't display the name of a patch when that patch is loaded. The copy-protection scheme gives you the option of installing a 50 MB file onto your computer or keeping the CD-ROM in your drive when you use the plug-in.

Native Instruments prides itself on the accuracy of Pro-Five's emulation, "right down to the waveform used in the test tone." It's been many years since I had my hands on a Prophet-5, but from all recollections, this year's model has got it spot on.

GakStoar delta 1.1 (Win; $75). Lin Plug's GakStoar delta is a virtual analog synth that offers 13 waveforms for each of its four oscillators. The frequency of each oscillator is tied to the same envelope, but each oscillator can have its own unique amplitude envelope. Oscillators 1 and 3 can be used as modulators for oscillators 2 and 4, giving you some very flexible AM and FM options. Unlike many analog synths, GakStoar's oscillator output can be sent directly to the main outs without passing through the customary filter sections.

The filters can run in parallel or in series, and each has settings for type (highpass, lowpass, and bandpass configured at 6, 12, 24, and 36 dB per octave), cutoff, resonance, keyboard tracking, and routing.

A large panel in the center of the interface provides easy routing of up to 17 modulation sources to 17 destinations. You can also scale the modulation amount in this window. Four LFOs, each with a choice of seven waveshapes, can be applied to a destination. Attack and Decay times for the LFOs can be adjusted from 0 milliseconds to 10 seconds in 1 ms increments, but it's not really possible to employ every conceivable value within that range if you're using the mouse to adjust the knob. The LFOs can be synched to the tempo of your sequence.

GakStoar has a single Undo level, but even that is more than many of the other programs in this roundup. Its Copy function makes it easy to copy the settings of one oscillator or envelope and apply them to another. A very thorough manual explains all of the program's features and adds to the appeal of this attractive synth.

ES1 (Mac/Win; $99). Emagic has released a plug-in for its Logic Audio series of sequencers that has capabilities equal to the best of this bunch. ES1 is a 16-voice, single-oscillator synth with the unique ability to toggle between a graphic synth-panel view and a simple numeric display containing all of the synth's parameters (see Fig. 12). This dual operating mode allows you to choose between a user interface optimized for live performance and one best suited to offline tweaking. Up to eight instances of ES1 can be run in parallel, provided you have the computing resources to tackle that task.

The single oscillator spans a 5-octave range and offers triangle, sawtooth, and variable-pulse waveforms. A suboscillator, which can be toggled on and off, tracks the pitch of the main oscillator and can be assigned its own waveform or noise. The LFO offers six fixed waveshapes, including a slowly evolving random waveform. The LFO can also sync to the tempo of your MIDI tracks; you can choose a number of divisions and a multiple of the tempo, or it can be freely set within a range of 0 to 24.5 Hz.

ES1's filter can be switched to four different rolloff settings: 12 dB, 24 dB, 24 dB Classic, and 24 dB Fat. (The Fat setting boosts the low end that's lost when resonance is used; Classic does not perform that compensation.) Filter parameters include Cut-Off, Drive (overdrive), Resonance, and Key (keyboard sensitivity), all of which can be modulated. You can also route an audio track through ES1's filter, a feature provided by a number of early synths, including the Minimoog Model D. Because of ES1's integration within Logic Audio, you can automate most other parameters, and Logic's own effects plug-ins can be applied to ES1's output.

A single envelope generator (ADSR) can be assigned to most parameters, though the routing method isn't as intuitive as it could be. You can also use the AGateR setting to deactivate the Decay and Sustain segments, or you can enable the GateR option to force the Attack time to 0 seconds. That leaves only the Release segment in effect.

ES1 ships with a number of presets, many of which emphasize the lower depths of the audio spectrum. "FM Bass," for example, has a very rich low end, and "Dry Bass," "DB Bass," and "Beyond Bass" also move an enormous amount of air. (Try opening the filter on some of the bass presets to give them a nasty bite.) The snare, hi-hat, and toms aren't quite up to the standards of some of the other drum presets in this roundup, but you won't limit yourself to presets when you have this much horsepower on hand.

RELEASE SEGMENTDesktop musicians today have an awesome amount of computing power available to them, and software synthesizers offer a great way to use that power. Unlike most of your synth hardware, soft synths improve with age as you put newer and faster processors at their disposal.

I hope you're inspired to add one or more of these programs to your audio toolkit, and I'm sure you'll find that many of them can add considerable new sonic possibilities to your rig. (To purchase or learn more about the products in this article, see the table "Soft Sites.") Soft synths are one of the most exciting new categories of music software to come along in recent memory, and you can bet they'll remain a major focus for creative manufacturers and developers for years to come.

EM associate editor Dennis Miller now has more software on his computer than at any time in history.

Though all of the programs I've covered here display their features on graphical user interfaces, that's not the only approach to generating sound in real time. Two sound-programming languages, Barry Vercoe's Csound (free) and James McCartney's SuperCollider ($250), also have the ability to create and modify sound as it plays back. Both programs let you send MIDI data from a controller to manipulate sound parameters, and SuperCollider can also incorporate graphic control panels containing sliders or buttons into its patches (see Fig. A).

With programming languages such as these, you build "virtual instruments" by accessing the huge libraries of functions that each language provides. In most cases, this involves typing text into a word processor or text editor (one is included with SuperCollider) and then telling the programs to compile or "render" the source files that contain your data. The end result is typically an audio file written to your hard drive, but your instruments might be designed with parameters that require real-time MIDI input. You can then play these instruments as you would any software synthesizer.

The real advantage to using this type of environment for synthesizing sound is the enormous power that both programs provide. There's no real limit to the type or complexity of patches that you can build; both languages offer significantly more building blocks than any of the programs reviewed in this roundup. Of course, as with any synthesizer, the more complex the sound, the less polyphony you're likely to get. And though many new "helper applications" are available (for Csound in particular) to make coding your sounds easier, not everyone will find composing with a word processor to be suited to their taste.

If you've never heard the music that sound-programming languages can create, you're in for a treat. Download Csound or the demo version of SuperCollider, and check out the wide range of styles in the examples included with both.

Many factors influence the performance of a soft synth on your desktop, but after surveying a number of the companies mentioned here, I can narrow the issue to several key points. First is the type and speed of the processor that you use. Most programs simply can't provide adequate performance on a slower machine, though some companies-Seer Systems, for example-have written their code in Assembly language, which allows them to get a little extra power from a slow computer. Plan on an absolute bare minimum of a Pentium II/200 MHz and a Mac PPC 604e/266 MHz, which are, according to the companies I contacted, the average minimum requirements. Just remember, however, that more is better.

Sound cards are the next major issue. Here, the quality of your card's drivers is key. Keep your eyes on the manufacturer's Web site for driver updates; you'll be surprised at how often they appear, especially for newer cards. Some soft-synth developers have worked closely with sound- card manufacturers to ensure optimal performance between the application and the hardware. Though this is more common with software samplers, it's a good sign when the hardware and software companies have been in touch with one another.

In the Windows world, using DirectSound improves on the performance that you would get from basic MME audio drivers, and all of the Windows synths support this protocol. The Mac has numerous audio protocols, and you'll have to determine which ones your synth and hardware are compatible with. For example, using ASIO can improve performance on any platform; I switched from MME to ASIO for the E-mu APS card on my PC, and latency in one of the synths went from a completely unacceptable 400 ms to 20 ms. But ASIO support in particular must come from both the sound card and the software; you may not have the right combination of the two.

The next issue is RAM. How much is enough? No one can say for sure, but one of the sequencers that I used on the Mac wouldn't even load a synth plug-in until I bumped up its allocation to more than 50 MB. All manufacturers agree that synths are RAM hungry, so like they say at holiday time, give and give generously. On the Mac, just be sure that you leave enough RAM for your computer to manage its other tasks.

There are dozens of other tweaks that you can use to eke out the best performance from a soft synth, such as turning off Virtual Memory on the Mac and loading only those extensions and control panels that you really need during a work session. (Check Extension Manager to find out which extensions are loading, or buy a copy of Conflict Catcher for even easier extension control.) PC users should shut down Autonotification in their Device Manager, because it periodically checks to see if a CD has been inserted into your drive, which could happen at just the wrong moment.

Today, having a powerful computer is a prerequisite for top performance; but there are many other considerations, including your audio hardware, the audio protocols that your platform supports, and the options that the software itself provides. No single solution works in all cases, but with a few tweaks (and a little luck), you should be all set.