You are here

Camel Audio Cameleon 5000

Resynthesizing Additive Soft Synth By Sam Inglis
Published April 2004

Cameleon 5000's Morph page. You can drag a dot around the Morph Square to perform real-time morphs between the four contributing Voice programs, which reside at the corners — or, as shown, you can set up complex envelopes which are triggered whenever you press a key. Shading from grey to white in the Morph Square shows how your morph evolves, whilst the conventional envelope editor above plots X or Y value against time.Cameleon 5000's Morph page. You can drag a dot around the Morph Square to perform real-time morphs between the four contributing Voice programs, which reside at the corners — or, as shown, you can set up complex envelopes which are triggered whenever you press a key. Shading from grey to white in the Morph Square shows how your morph evolves, whilst the conventional envelope editor above plots X or Y value against time.

Resynthesis promises to combine the realism of a sampler with the flexibility of a synth. A great idea, but no-one has made it work — until now...

If I had to compile a 'memorable products of last year' list, I would definitely reserve a place for Virsyn's Cube software synth. With Cube, which I reviewed in SOS September 2003, Virsyn made the powerful but neglected technique of additive synthesis truly accessible for the first time. You could create fascinating new sounds by manipulating the level and pan of every individual partial, and you could draw on a useful range of preset harmonic spectra derived from real instruments and human voices. Cube was a very powerful and original synth; but what it lacked was the ability to analyse an input sound automatically, work out its harmonic content, and translate that information into an additive synth patch.

Resynthesis, to give this technique its proper name, has been the ultimate goal for the developers of additive synths since the days of the Fairlight Computer Musical Instrument, promising to combine the realism of sampling with radical freedom to repitch, morph and generally monkey around with the sound. Even last year, I suspected that this might be too ambitious for current technology, but things move fast in the world of soft synths, and only six months later I find myself sat in front of another additive synthesizer — one that promises fast, powerful and accurate resynthesis of sounds based on user samples. Cameleon 5000 is the first synth to emerge from Ben Gillett's Camel Audio stable (do camels live in stables?), and runs as a VST Instrument under Windows and Mac OS X, with an Audio Units version also available on the latter platform.

Presets & Programs

If you set aside the resynthesis and some stylish virtual walnut panelling, Cameleon 5000 actually bears a fairly close resemblance to Virsyn Cube. As with Cube, the basic structure of a Cameleon patch consists of four separate additive sound sources, which are visualised as being at the corners of a square. And as with Cube, you can use MIDI controllers and complex multi-stage envelopes to morph between these four sources. This allows you to create lengthy, evolving patches, or hybrid sounds which seamlessly combine, say, the attack portion of a duck's quacking with the sustain of a saxophone and the decay of a marimba. The morphed output of the four additive sources is then fed through an effects section comprising a formant filter with up to 128 breakpoints, distortion, a conventional resonant filter, chorus, stereo delay and reverb.

Like Virsyn, Camel Audio have realised that additive synthesis can be daunting for the newcomer, and they've tried to make it possible to get a lot out of Cameleon without getting too deeply involved in editing the individual additive sources. There are, in effect, two 'levels' of Cameleon patch: Voice Programs, which contain all the parameters associated with an individual source, and Presets, which assign different Voice Programs to the four sources and store global information such as morph envelopes, modulation and effects settings. Voice Programs are independent of Presets, and a single Voice Program can be used in different Presets (or even more than once within the same Preset). Presets are, in turn, arranged in banks. The Cameleon interface is divided into a total of eight editing pages; four deal with the parameters for Voice Programs A, B, C and D, while the other four address the global parameters that make up a Preset.

A browse of the factory sound banks reveals that Camel Audio have focused far more than Virsyn on recreating real instruments, showing off their resynthesis technology. The Preset that's loaded by default, for instance, is an acoustic bass every bit as good as your average workstation S&S sound — but no samples are involved in playing back this sound. Some of the other 'real instruments' are even more impressive, and with Presets like Celesta, Finger Bass and Steel Drums, it's only the utter lack of multisample split points and aliasing artifacts that tell you this is not a sampler. Others, such as Oboe and Alto Sax, are much wider of the mark, but at the very least, the tonal quality of the real instrument always comes through. Of course, resynthesis isn't restricted to samples of acoustic instruments, and Camel Audio have also applied it to classic analogue and FM synths such as the Minimoog and DX7, often with excellent results. There are also numerous excellent sound effects, along with a small bank of percussion Presets, though this is clearly not one of Cameleon 's strengths.

What Is Additive Synthesis?

Mathematician Jean-Baptiste Fourier demonstrated that any repeating waveform can be reduced to a series of sine waves of different frequencies and amplitudes, superimposed on top of one another. When that waveform has a definite musical pitch, it's because the frequencies of the contributing sine waves are mathematically related, and it's this insight that inspires additive synthesis. A sampler simply records and plays back complex waveforms directly, without attempting to analyse them. Additive synthesis, by contrast, 'builds up' a waveform by adding together sine waves, beginning with the root or fundamental frequency of a note and working up through the appropriate related frequencies (known as harmonics or partials). In theory, this technique promises sounds that are realistic, yet malleable in ways that samples can never be.

Of course, it's not quite as simple as that. In order to provide a realistic recreations of real instruments, an additive synthesizer also needs some way of reproducing the unpitched noises, like the 'thunk' of a piano hammer or the 'chiff' of a flute, that they make in addition to their harmonically related partials. Moreover, the harmonics produced by real, physical systems don't always have precisely the relation to the fundamental frequency that theory says they should, and these discrepancies must be accounted for. Most importantly of all, an additive synthesizer needs to do more than just reproduce a 'snapshot' of an instrument's waveform: it needs to mimic the way that waveform changes over time. As most devotees of SOS 's Synth Secrets series know only too well, in most real-world sounds, the higher-frequency partials die away more quickly than the fundamental and the lower harmonics. The basic task of reproducing the harmonic content of a source sound can require 64 or even more partials, and if the sound is to evolve naturally rather than remain static, each of these must have its own envelope. Reproducing the real-world detuning of the partials and unpitched content of an instrument's sound is also no trivial matter.

Implementing all of these features in a digital additive synth demands a lot of processing power, which is one reason why additive synthesis has failed to dominate the synth market. The '80s saw some bold attempts, from the original Fairlight Page 4 to the Technox Axel, but none had the processing power available to reproduce real-world sounds in a convincing fashion, and the additive dream disappeared under a rash of affordable, sample-based synths.

It resurfaced again in 1997 in the shape of the Kawai K5000. This sleek silver beast introduced a number of innovations, perhaps the most notable being that you could combine PCM samples with the additive oscillators to reproduce the noisy and enharmonic elements of a sound. With 64 partials per oscillator, each with its own amplitude envelope, and a sophisticated 128-band formant filter, the K5000 was a far more capable instrument than any of its additive ancestors. Yet the way its power was presented highlighted another key problem with additive synthesizers: the more powerful they are, the more parameters the user has to adjust in order to create a sound. The K5000 was a pig to program, even with the aid of a big LCD and a specially adapted version of Emagic's Sound Diver, and it was not a commercial success.

There is one obvious way to resolve the 'power versus programmability' dilemma. Creating a sound from scratch in a synth that offers as many parameters as the K5000 is a daunting task. But what if that synth could do the programming work for you, analysing an input sound, working out how its harmonic structure changes over time, and setting its own synth parameters appropriately? This technique, known as resynthesis, has long been the ultimate goal of additive synth designers, but it poses a formidable technical challenge. Both the K5000 and Virsyn's Cube soft synth included additive templates derived from real-world systems such as plucked and bowed strings, but the breakthrough in Cameleon 5000 is its ability to analyse and recreate both the pitched and unpitched components of a sample. This is a huge development, which makes it possible to access the power of additive synthesis without getting involved in editing hundreds of parameters.

Meet Morph

So far, there's little to separate Cameleon from more conventional S&S or 'virtual-analogue' software synths. However, that's because all the Presets I've mentioned so far only use a single Voice Program, loaded into additive source A. The editing page that's presented by default when Cameleon opens up is the Morph window, and it's here that things begin to get interesting. It's the work of seconds to call up additional Voice Programs into sources B, C or D, and this is the first step towards creating your unique blend of instruments. You can then create a hybrid sound by picking up the white dot with the mouse and moving it to a different position within the Morph Square. Movements affect the sound in real time, so you can use your sequencer's automation to record hands-on morphs, and you could also assign the position of this dot to a suitable X-Y controller.

Alternatively, you can create complex morphs that are stored as part of the Preset. Clicking the Morph Timeline button hands over control of the morphing to two envelopes, which you can edit in the window above the Morph Square. The X envelope, as you'd expect, controls the morph balance between Voice Programs A and C on the left of the square, and B and D on the right, while the Y envelope adjusts the top-to-bottom morph balance. Editing the envelopes is a simple matter of clicking to create breakpoints and dragging them to the right positions, and you also have the option of dragging left and right loop point locators to define a part of the envelope that should loop while a note is sustained. As of version 1.2, the Morph Square provides another way of visualising and editing your morph envelope, with the breakpoints plotted in the appropriate positions and the direction of the morph indicated by a shading from grey to white.

So far, so identical to Virsyn Cube, and the similarities continue. Clicking in the time ruler for the envelope editor and dragging the mouse up or down zooms horizontally in and out, while hitting the Sync button switches the timescale to bars and beats, locked to the tempo of Cameleon 's host. This allows you to create sounds that morph rhythmically in time with your song — and since there's no sample-slicing or time-stretching involved, there's no 'root' tempo for a Preset, and no restrictions on how far you can vary it. Another handy feature added in version 1.2 is the ability to load and save Morph Timeline envelopes independently of Cameleon Presets.

The X-Y morphing envelope can be sync'ed to host tempo, for rhythmic effects.The X-Y morphing envelope can be sync'ed to host tempo, for rhythmic effects. Cameleon 's morphing does differ from Cube 's in a few significant ways. The most important is that Cameleon treats each Voice Program as being made up of two separate elements: a pitched harmonic content and an unpitched noise content. When the Timeline is active, a single X-Y enveloped controls both morph elements. When it's switched off, however, four buttons at the right-hand side of the Morph page come into play. These are labelled All, Harmonics, Noise and Amplitude, and only one can be selected at once. If that button is All, a white dot appears in the Morph Square, and moving this around morphs all parameters simultaneously. If you click one of the other buttons, the white dot divides into separate coloured dots, allowing you to set individual morph settings for the harmonic content, noise content or overall amplitude envelope. There's also a Random Morph option, accessed by clicking on the camel icon (obviously!), which chooses a Morph Mode and positions the dots at random. Whether the Morph Timeline is switched on or not, the X and Y values for Harmonics, Noise and Amplitude morphing are always available as separate modulation destinations in the Mod page.

In normal use, it's rare that you need the fine control provided by separate Harmonics and Noise morphing, and All nearly always produces the most interesting results; restricting the morph to the noise content alone can be subtle to the point of inaudibility. I was slightly surprised to find that, although some of the factory Presets are designed to show off Cameleon 's morphing capabilities, few of them load with the Morph Timeline active, or with any X or Y morph envelopes defined.

Morphing In Action

The most important thing to say about Cameleon 's morphing is that it is easy to use and offers a vast array of creative possibilities. Since only one Voice Program is required to reproduce any source sound, you can achieve startling results simply by loading two different 'real instrument' Voice Programs into sources A and B, and using a simple linear morph to fade between them. You want an instrument with the sharp attack of a picked banjo, but the body and sustain of a bowed cello? No problem. A French horn that doesn't die away, but slowly evolves into a sinister synth pad? Right here. A barrelhouse accordian that turns into a celestial choir? At your service. The morphing is just as impressively smooth as it was in Virsyn Cube, and with a total of four additive sources to play with, lengthy, evolving soundscapes are just a couple of mouse-clicks away.

I encountered a few problems with morphing in early versions of Cameleon, but the v1.2 update ironed them all out with impressive speed! If I could register a couple of wishes for future versions, they would include rubber-banding — in other words, the ability to select a section of the envelope containing multiple breakpoints, in order to move them all at once. Similarly, it would be great to be able to select a section of the envelope and scale it up or down in size, both vertically and horizontally. Cameleon might also benefit from making it easier to incorporate rhythmic elements into a morph. Because all four Voice Programs in a Preset are triggered simultaneously when you press a key, you usually end up morphing from sustained sound to sustained sound. If it was possible to either delay the triggering of the Voice Programs in the Morph page, or to retrigger them at will, it would be easier to build into a morph the kind of transient attack sounds that are needed to create a rhythmic loop. It is possible, as we'll see, to edit individual Voice Programs to give similar results, but it would be more convenient and more flexible if this could be done in the Morph editor. For one thing, the Morph envelopes can be sync'ed to host tempo, whereas the Voice Programs' own envelopes cannot.

Global Warming

Apart from the Morph page, there are three other editing pages devoted to the global parameters that make up a Preset. The Effects parameters are, for the most part, self-explanatory (see screenshot on page 67), and the only thing you wouldn't find in a standard subtractive synth is the Formant Filter. This can be thought of as a complex EQ spanning the entire frequency spectrum, with up to 128 bands that are created by clicking and dragging. This marks another difference between Cameleon and Cube — the latter supplied a separate formant filter for each oscillator, with lists of preset curves modelled on real-world resonant structures such as the human chest and larynx.

The Easy page's Stretch function allows you to make huge changes to the length of a sound without any of the usual time-stretching artifacts.The Easy page's Stretch function allows you to make huge changes to the length of a sound without any of the usual time-stretching artifacts.Camel Audio haven't implemented any of the clever drawing modes that Virsyn came up with for creating comb responses and the like, and it's harder work to create a complex filter shape in Cameleon. As we'll see, however, this is forgivable because it's perfectly possible to use Cameleon without using the formant filter at all. You can treat it as a key part of an additive sound if you want, as Cube and the Kawai K5000 do, but you don't have to. Saving and loading of formant filter shapes is another welcome addition in v1.2 of Cameleon, meaning that you only have to do the hard work of creating them once. A nice bonus compared with Cube is that Cameleon 's filter response can be modulated by any source of your choosing, in effect sliding the entire filter curve to the left or right along the graph, although there's no visual feedback to show what is happening.

The other Cameleon effects are more familiar. They begin with a compression and distortion unit based on Camel Audio's popular Camelphat plug-in, continue with a conventional multi-mode, envelope-following resonant filter also derived from Camelphat, and proceed through a chorus unit and a tempo-sync'ed stereo delay (with delay time always set in bars and beats rather than seconds), before finally encountering a basic but perfectly acceptable reverb. Once again, clicking on the camel generates random settings. In general I found the chorus and delay the most useful of the effects; the distortion, compression and resonant filtering all work well enough, but tend to obscure rather than emphasise the individuality of Cameleon 's sounds.

A Cameleon Preset also encompasses some global synthesis parameters, and these are presented collectively in the Easy page. Three Timbre controls allow you to make quick-and-dirty adjustments to the entire sound, affecting all of the additive sources: the Brightness control makes the sound brighter or darker not by EQ'ing it, but by increasing or decreasing the amplitude of the higher-order partials, while Harmonics and Noise balance the relative levels of the pitched and unpitched components of the additive sound engine. There's also a basic, tempo-sync'ed LFO which can be routed to pitch or amplitude modulation, plus a portamento option and a Stereo Width control. The latter doesn't compare to Cube 's ability to pan every partial individually, but is much simpler to use, and does make a noticeable difference to the stereo spread of a sound.

The Easy page (shown on page 68) also provides a global amplitude envelope. This is pretty basic, but as we'll see, Cameleon 's individual Voice Programs include sophisticated envelope generators, and a complicated global envelope would be unnecessary. Instead, you get conventional Attack and Release controls, and a Stretch control. Set to 100 percent, this has no effect; set either side, it lengthens or shortens the amplitude envelope for every Cameleon partial. The results are rather like time-compressing or time-stretching a sample — except that Stretch doesn't introduce any unpleasant artifacts at all. That's quite impressive when you consider that its range runs from 20 to 500 percent! Assigning Stretch amount as a modulation destination yields some interesting expressive possibilities. Stretch has no effect on the X or Y morphing envelopes.

Cameleon is a pretty CPU-intensive synth, and you can restrict its processing load in two ways: by reducing the polyphony from its maximum of 12 notes per instance, or by limiting the number of partials that it uses to create its sounds. In many cases, the partials towards the top of a sound's frequency spectrum are very quiet compared to those closer to the fundamental, and you can often dispose of them without noticeably changing the sound. Cameleon 's CPU load also spikes briefly at the start of a note, and I found this was often enough to cause a click on the basic (700Mhz G4) iMac I used to test it. If you have the freedom to adjust your soundcard latency, increasing the buffer size (and hence the latency) apparently helps reduce this spike.

The final global editing page is Mod (shown opposite), which allows you to assign up to eight source-destination routings, and offers a couple of additional tempo-sync'ed LFOs, with their own envelopem to act as modulation sources. One source can modulate multiple destinations, and a nice touch is that you can restrict the range over which a source acts, for instance so that the entire travel of the mod wheel only introduces a small amount of vibrato. By setting the Minimum slider higher than the Maximum one you can even have modulators act in reverse, so that pushing the mod wheel up progressively reduces vibrato. This all works very well, provided the Cameleon parameter you're controlling is set at source within the range you've chosen.

The Additive Engine

If an additive synth is to stand any chance of reproducing realistic sounds, it needs to be capable of reproducing the amplitude envelopes and detuning information for a decent number of partials — most developers seem to agree that at least 64 are needed — while a complex noise generator is needed to create the unpitched element of the sound. In Virsyn's Cube and the Kawai K5000, formant filtering also formed a key part of the additive synthesis process, the basic idea in both cases being similar to physical modelling synthesis: the additive oscillator itself was used to reconstruct a source signal (such as a plucked string), while the formant filter could model the behaviour of a resonant body (such as a guitar). This division of labour has some advantages, perhaps the most important being that it can reflect the ways in which a real instrument responds to varying pitch or loudness — the source signal can be pitched up and down while keeping the response of the resonant body fixed, which is of course exactly what happens in a real instrument like the guitar.

I Can Sing A Rainbow

New in version 1.2 of Cameleon 5000 is the ability to import bitmapped graphics files (with a '.BMP' extender) and turn them into Cameleon patches. Importing a BMP file seems to produce a harmonic pattern and envelope, but no noise element or detuning. The process is near-instant — much faster than that of analysing a sample — and can yield some interesting results; most of the graphics files I tried produced burbling or chattering sounds that evolved over a few seconds. It's not always easy to predict how a given picture will translate into sound, but if you're ever stuck for inspiration, this provides a quick and easy way to get things moving on the programming front.

It is possible to do exactly the same in Cameleon, thanks to the global formant filter accessed the Effects page. However, this approach has its drawbacks if you want to resynthesize a sound from a single sample. The reason for this is that if all you have is a sample of the real instrument, it's impossible to isolate the individual contributions of the source signal and the resonant body. If there's, say, a peak at the seventh harmonic, it's hard to know whether that's due to the frequency content supplied by the plucked string, or the way the instrument's body resonates. Cameleon 's aim is to offer true resynthesis from single samples, and it takes a rather different approach. Individual Voice Programs use formant filtering only to shape the noise element of the sound, but there's much greater control available at source over the harmonic and noise components. Where other additive synths such as Cube tend to provide only a simple AD envelope for each partial, each Cameleon 5000 Voice Program has its own multi-stage envelope with up to 128 breakpoints. Not only is the amplitude of each of the 64 partials editable at each stage, but also its degree of detuning away from the frequency that theory says that partial should have.

Likewise, Cameleon 's noise generators are more sophisticated than those available in other additive synths. Cube, for instance, generated its noise by modulating the harmonic partials, meaning that it always followed the same simple A-D envelopes they did, and followed the pitch of each note, usually ending up with broadly the same frequency 'colour'. Cameleon, on the other hand, takes raw white noise as its starting point, and allows you to shape it using a formant filter (paragraphic EQ) just like the global one in the Effects section. Again, this can have up to 128 breakpoints, allowing you to create sophisticated shapes; and like the harmonics, the noise content has its own envelope. Up to 128 stages are available, each storing a complete snapshot of the noise shape, so it's possible to create very complex noise patterns which are completely unrelated to the harmonic elements generated by the Voice Program.

From The Camel's Mouth

With Cameleon 5000, Camel Audio's designer and lead developer Ben Gillett aimed to implement true resynthesis, a goal that has eluded synth designers for more than 20 years. I asked him a few questions about how he'd gone about it.

What got you interested in the idea of resynthesis?

"My interest in resynthesis came from two independent angles. Firstly, I'm a musician, and I've always sought ways to create and manipulate sounds in new ways. Resynthesis offers the ability to manipulate sounds in ways that are completely impossible with conventional sampling. For example, being able to combine the amplitude of one sound with the harmonics of another, and the noise of another, was a very appealing idea to me, since it offers an intuitive way to design sounds.

"Before beginning work on Cameleon, I was doing research at Edinburgh University into voice transformation — a little like TC's Voice Prism, but where you simply give the system an example of the target speech which it should map to. I had a number of papers published on the system which I created — if any readers want to go and look it up! During this research, I made use of something called the harmonic-plus-noise model of speech. I thought that it would be pretty cool to use this to create something which musicians could use to design new sounds. I'm proud to say that Cameleon is the first synthesizer to make use of this harmonic-plus-noise research."

These days, conventional software samplers are extremely advanced. Are there still advantages for resynthesis over sample-based techniques?

"Sample-based techniques have had a great deal of research directed towards them. However, there are many things that cannot be done with sample-based techniques. With resynthesis, you have far more control over the sound — you can take a sound and then edit the amplitudes of the harmonics in any way you wish. With harmonic-plus-noise resynthesis, you can separately alter the harmonic and noise components — for much greater control over sounds. Also, resynthesis allows you to morph between sounds, rather than simply crossfading. Morphing offers a great deal, in terms of opportunities for creating hybrid instruments, and evolving sounds.

"Also, Cameleon doesn't suffer from the side-effects of sample-based pitch-shifting. Firstly, the separation between noise and harmonic spectra in combination with keeping the noise component constant whilst shifting the harmonic component leads to better pitch-shifting. This models most instruments and voices well; for example, when a speaker increases the pitch of their voice, the noise part remains pretty constant regardless of the frequency of the speech. Secondly, because of the way the analysis works — in terms of breaking the sound down into a series of sine waves, it is possible to play the sound back at the same speed, regardless of the pitch of note. This also offers creative possibilities, through the ability to, for example, take a piano note, and stretch it into a long evolving sound."

Additive synthesizers have often been called 'cold' and 'digital-sounding', but that's certainly not true of Cameleon. How have you avoided this?

"In theory, pitched sounds are made from a fundamental frequency, and partials whose frequencies are whole number multiples of the fundamental. However, for real instruments, things are a bit more complicated than that — the partials aren't perfect integer multiples of the fundmental, but are instead a little lower or higher in pitch, and this will also vary over the course of the note. Part of the reason why additive synthesis has previously been regarded as being cold, digital or boring is precisely because the partials cannot be detuned appropriately. In fact, you can test the effect of not having detuning of partials by loading up Cameleon, picking a sound you are interested in, and turning the Harmonize dial all the way up — this sets all the detuning to zero. On some sounds the effect is not noticeable, whereas on others it makes the sound a lot less interesting and more synthetic-sounding. For example, get a low note on a piano and try turning Harmonize on — the resulting sound will be far more synthetic.

"The other reason why some other additive synthesizers sound relatively 'digital' is that they typically have only an AD or ADSR envelope to control the partials, rather than a 128-point breakpoint envelope per partial, as Cameleon has. Also, Cameleon is the only additive synthesizer to accurately reproduce the noise component of sounds."

Can you explain how the process of analysing samples actually works?

"Basically the program analyses the sound as a spectrogram — like that which you can display in a sound-editing program — with time going across the x-axis, and frequency going up the y-axis. It then looks for partials — ie. bright lines on a spectrogram — and tracks these. Once it has found all those it can, it then tries to work out what the fundamental frequency (pitch) of the sound is, and then allocates the partials to each harmonic. Cameleon then assumes that the rest of the sound must be the noise part and figures out appropriate settings for the noise generator. The full implementation is much more complicated, but that is the basic principle."

Editing Voice Programs

With two 128-stage envelopes, each stage of which stores hundreds of parameters, Cameleon 's sound engine definitely has the potential to intimidate the user; and like Virsyn, Camel Audio have tried to develop innovative editing features to make the interface easier to navigate. The editing page for each Voice Program (shown on page 71) is based around parallel graphical displays for the Harmonics and Noise components of the sound. In each case, the lower one displays the envelope for that component, with breakpoints added by clicking and removed by shift-clicking. The height of each breakpoint controls the overall level of the Harmonics or Noise at that point. If you have Loop selected in the Easy page's Amplitude envelope, the left and right loop markers will also appear.

For the Harmonics generator, the upper half of each display can be switched to represent either the amplitude or detune amount for the individual partials, and can be zoomed to show only the lowest (and most important) 32 partials if you wish. Various shortcuts allow you to select multiple related partials for simultaneous editing, such as fifths, octaves, even and odd harmonics, while a Harm(onise) dial allows you to progressively over-ride any detuning to return the partials to their theoretically perfect pitches. The upper half of the Noise generator's display is exactly like the Formant Filter on the Effects page, and is shaped by adding, removing and dragging breakpoints in the same way.

In both cases, two editing modes are available. In Breakpoint mode, you select one breakpoint on the Harmonics or Noise envelope, and the upper display shows the partial or noise-shape settings only for that particular stage in the envelope. In Overall mode, by contrast, it's as though all the breakpoints in the Harmonics or Noise envelope are selected simultaneously. What you see in the upper display is an amalgam of the settings for each stage in the envelope. The idea is that Overall mode allows you to make broad-brush adjustments, such as raising or lowering the amount of third-harmonic content in the entire sound or trimming some high frequencies from its noise element, while Breakpoint mode is used for detailed editing. For instance, it's characteristic of some acoustic instruments for the attack phase of the sound to be sharp or flat compared to its sustained component, and to reproduce this, you'd need to use Breakpoint mode to adjust the detuning for the initial stages of the Harmonics envelope only.

Three different editing behaviours can be selected. In Fix mode, moving the currently selected breakpoint leaves the others unaffected. In Slide mode, moving a breakpoint to the left or right causes all the breakpoints to the right of it to follow suit, while Str(etch) mode does the same, except that moving a breakpoint also causes those to the left of it to squeeze together or spread out accordingly. These are a big help, so it's a shame that they don't apply to the breakpoints in the noise spectrum or the envelopes in the Morph page. Sliding all but the first breakpoints to the right allows you to, in effect, delay the attack phase of a sound, which goes some way to making up for the lack of this feature in the Morph page.

As of Cameleon v1.2 there are three sets of Load and Save buttons in each Voice Program edit page. Those at the top left allow you to store entire Voice Programs, including both harmonic and noise data, while two additional sets permit the independent loading and saving of harmonic and noise envelopes — should you wish to combine, say, the harmonic spectrum of a violin with the noise output of a human soprano. Each component now also boasts two Paste buttons, the leftmost of which ought to be named Copy. These allow you to copy and paste individual breakpoints in the envelope, which is useful when you're creating a sound with a looping element, as you can precisely match the harmonics or noise levels at the start and end of the loop. There are also volume controls allowing you to adjust the overall level of harmonics and noise in the sound.

Overall, Cameleon 's Voice Program editor does a pretty good job of making its power accessible, though I didn't find it quite as quick or easy to use as its equivalent in Virsyn Cube. This is partly because there are many more parameters to edit, and partly because shaping the noise spectrum using breakpoints isn't as fast as simply drawing with the mouse.

It would, nevertheless, be perfectly possible to create interesting Cameleon patches from scratch using the Voice Program editor. However, Cameleon has an ace up its sleeve which means most users probably won't ever want to do so: it can take a WAV or AIFF sample, analyse its contents and turn the results into a Voice Program.


I've been saving Cameleon 's best feature until last, because its resynthesis capabilities are truly special. Clicking the Import button brings up a file browser, which you can point at any 16-bit WAV or AIFF file. In general, the qualities that make for good multisamples are also appreciated by Cameleon, so your source files should be cropped at start and finish, normalised to 0dBFS and correctly tuned. They should also be monophonic and free from pitch changes, spill or excessive reverb.

The Voice Program editor window. Sounds are divided into their pitched components (the top two windows, under the heading 'Harmonics') and their unpitched, or noise components (below, labelled 'Noise'). In both sections, the 'Amp' curves trace the overall amplifier envelope of that component over time. The uppermost display here shows the amplitude of the individual harmonics that make up the pitched component of the sound, while the third curve, labelled 'Frequency', shows the amplitude of the frequencies that make up the unpitched component of the sound over the range of audible audio. From the curve shown here, for example, you can see that most of the frequencies in the unpitched component of the sound are down at the low end of the audible spectrum, near 20Hz, and also between 3.3 and 5kHz.The Voice Program editor window. Sounds are divided into their pitched components (the top two windows, under the heading 'Harmonics') and their unpitched, or noise components (below, labelled 'Noise'). In both sections, the 'Amp' curves trace the overall amplifier envelope of that component over time. The uppermost display here shows the amplitude of the individual harmonics that make up the pitched component of the sound, while the third curve, labelled 'Frequency', shows the amplitude of the frequencies that make up the unpitched component of the sound over the range of audible audio. From the curve shown here, for example, you can see that most of the frequencies in the unpitched component of the sound are down at the low end of the audible spectrum, near 20Hz, and also between 3.3 and 5kHz. Cameleon is pretty good at guessing the root note of pitched samples, but if root note information is stored in the WAV or AIFF file, it will use this instead of its own guess. Some commercial sample libraries include WAV files containing incorrect MIDI note information, and this will tend to throw Cameleon 's import function wildly off the mark. This can sometimes generate interesting and unexpected sounds, but if you don't want it to happen, you'll need to amend the note information with a suitable audio editor; alternatively, you can specify a pitch and octave in the file name, which overrides any information in the file itself.

Once you've chosen your sample, Cameleon takes a few seconds to analyse it, and in most cases, you get a perfectly playable Voice Program without any further tweaking. What's more, the analogy with multisampling goes further. Most real instruments respond differently at different points in their range, or when played with more or less force; and it's impossible to extrapolate this from a single sample. To this end, each Cameleon Voice Program can actually contain the information extracted from up to 16 samples, comprising up to eight different pitches with one or two velocity layers. Each different pitched 'sample' is edited separately in the Voice Program editor, but you can choose whether to edit the Soft and Loud layers in linked or separate modes. If you're not using the velocity layering, a Soft Cut option allows you to link brightness to keyboard velocity, which can help to mimic the loudness-related changes in timbre exhibited by most real instruments.

This 'multisampling' capability might not sound that impressive compared to those vast Gigastudio libraries that sample every note of the concert grand at a zillion velocity levels, but to think that is to miss the point. Cameleon can pitch a single 'sample' up and down the keyboard without any of the 'chipmunking' or other pitch-shifting artifacts that afflict samplers, so there's really no need to sample every note of an instrument. Just as Cameleon can morph smoothly between two different Voice Programs, moreover, so it can also interpolate evenly between pitched 'samples' that are an octave or more apart, eliminating those awkward jumps in timbre as you play up and down the keyboard. Nor will you hear nasty steps between loud and quiet samples.

There are a few areas where the interface could be tightened up: it's a shame there's no Preview function in the Import panel, and where you're planning to import several samples to create a 'multisampled' patch, it would be more convenient if this could be done in one go. It would also be nice if there was some way to edit envelopes and so on for all the 'multisamples' in a Voice Program simultaneously. However, these are very minor complaints about what is, in essence, a killer feature. Provided I fed Cameleon a clean sample, the resynthesized result was impressively close to the original in the great majority of cases, and I was particularly impressed by the accuracy with which Cameleon reconstructed the harmonically complex sound of instruments like the sitar; even where the overtones in a sample were obviously not perfectly in tune with the fundamental, Cameleon coped very well. It only really fell down when I tried it out with completely enharmonic sounds such as cymbal crashes, and even then, the results were often interesting, even if not accurate.

A Voice Program derived from a single sample was normally playable within a two-octave range, where pitching the sample itself up or down would have created obvious artifacts, while only three or four WAV files are needed to create a 'multisampled' instrument that works across the entire keyboard. It's also worth pointing out that what Cameleon takes in terms of CPU horsepower (compared with a soft sampler) it gives back in disk space and memory usage: most Voice Programs are under a megabyte in size. Resynthesis as implemented here is a powerful and unique creative tool that's a doddle to use, and it really lifts Cameleon 5000 out of the ordinary.

Summing Up

It's not often that SOS devotes so much space to a software synth that's currently available only as an Internet download, but then it's not often that a soft synth makes as much of an impact as Cameleon 5000. It's one of those instruments that offers almost unlimited programming potential, whilst allowing the casual user to jump in after only a quick browse of the manual. In its resynthesis capability, it offers a killer feature that's easy to use and, at present, unmatched by any other synth. Resynthesis might turn out to be the coming thing — Disco DSP's Windows-only Vertigo software synth already implements a basic form of it, and at this year's Winter NAMM show, Virsyn announced that the forthcoming version 1.5 of Cube will add a resynthesis feature — but as far as I know, there's currently nothing else around that comes anywhere near the power and realism of Cameleon 's 'multisampling', harmonic-plus-noise design.

What matters above all, though, is that Cameleon is a hugely playable instrument with an unusually supple, fluid sound. Features such as morphing and Stretch add masses of expressive potential, and Cameleon 5000 is very much a software synth for musicians as well as sound designers. The pace of development at Camel Audio seems very swift, with the v1.2 update in particular representing a huge improvement. I can only recommend that you head to Camel Audio's web site as soon as possible and download the free demo version. I think you'll be impressed.

Test Spec

Cameleon 5000 versions reviewed: v1.1 & v1.2.
700MHz Apple iMac G4 with 256MB of RAM, running Mac OS v10.3.2.
Steinberg Cubase SX for Mac v2.01.


  • The first synth to offer genuine resynthesis.
  • It's easy to use and it sounds great.
  • A wealth of morphing and sound-shaping possibilities you won't find in any sampler.


  • Resynthesis doesn't always offer a carbon copy of the source sound.
  • Demands a lot from your computer.


It might not be too much of an exaggeration to describe Cameleon 5000 as a minor revolution in synthesis. It allows you to distil the essence of a real instrument and manipulate it in ways that simply haven't been possible until now; and most importantly, it's a playable and expressive instrument in its own right.


$199 or 159 Euros (around £105 at time of going to press).