Its manufacturers claim the use of Artificial Intelligence enables the Kurzweil to sound better than any other sampled-sound instrument, but can it really replicate all the timbral characteristics of a concert grand piano?

When the Kurzweil 250 made its first public appearance at the June 83 NAMM show in Chicago, it set a lot of tongues waggling and a lot of mouths salivating. A digital keyboard with high-quality sampled sounds that captured the dynamic timbral changes of a concert grand for under $10,000? Well, it couldn't be done. Could it..?

In fact Kurzweil have succeeded magnificently in casting aside all the carps of doubting Thomases, because what they've done works superbly well. But the 250 hasn't arrived without incurring a few penalty points along the way. For starters, when the 'under $10,000' price tag was first quoted, the pound was rather nearer the two-dollar mark than it is now, which means that a middling indentation into musicians' pockets has turned into an RRP further into the big league of computer music systems. As of October 22, both the Kurzweil's UK importers - Scenic Sounds and Syco - have agreed on an RRP of £10,995, and that's without VAT.

But let's look on the bright side first. Where the 250 really scores over its more expensive and world-wise competitors is in the way it approaches the technique of sound sampling. As I've said before, it's just about the easiest thing in the world to digitise sound into memory and then chuck it out again as a more or less pale imitation of the original. The only problem is that you rapidly find yourself on the receiving end of a reality that's doing its damnedest to make life difficult, and which goes something like as follows:

1 High bandwidth sampling requires very fast sampling.
2 Very fast sampling chews up memory like there's no tomorrow.
3 Memory (ROM especially) is both expensive and power greedy.
4 The finished product ends up being costly, heavy, and hot.

Taking the 360 Systems keyboard as a commercial example of this flow of sampling logic, we find 352K of eight-bit storage space (48 x 2764 ROMs) being used for just a single sound - the bowed violin. And even though this is a realistic bowed violin that decays quite naturally without using looping to prolong the sound artificially, it's still only a collection of four string samples taken at various intervals, and more to the point, only at a single dynamic level. If you're happy to pay £700 for a high-quality string orchestra that's stuck in a single dynamic groove, look no further.

Background

However, Ray Kurzweil, principal designer behind the 250, had other plans in mind for his keyboard. His aim was to create a system that allowed sounds to be digitised, analysed, coded, modified, and played, taking into account timbre changes that are both pitch- and amplitude-dependent. His starting point was the piano, it's the instrument that just about every manufacturer has sought to emulate, but few have come within spitting distance of reproducing its vast range of colours. Kurzweil puts the pianistic sampling situation like this:

...each of the piano's 88 keys can create about 250 distinct timbres. Each of these 22,000 sounds lasts about 25 seconds for a total of about half-a-million seconds. If these were recorded using standard digital audio disk techniques (44,000 16-bit samples), it would require about 350 billion bits. With 256Kbit memory chips, that would require over a million memory chips for the piano alone!

Faced with all this mind-boggling mathematics, a lesser man would have given up and joined the West Coast lemmings for a life of sun, sea, and silicon dreams. Not so Ray Kurzweil. Fact was, he had another string to his bow in the form of the work his other company, Kurzweil Computer Products, had put into a machine called the Kurzweil Reading Machine. This is, basically, a $30,000 box of tricks that translates printed text into phonetic speech (Stevie Wonder has one of these as well as his 250 Keyboard). What's important about the Reading Machine is that it uses the principle of Artificial Intelligence to learn as it goes, rather than just diving into the deep end every time a particular character is encountered. Kurzweil realised that a similar technique could be applied to sound, so that if a sound changes its timbre as it's played louder or softer, a computer can be used to analyse these changes and work out the rhyme and reason - the timbral algorithm, if you like (or what Kurzweil themselves call the 'Contoured Sound Modelling') - needed to recreate a particular sound at a given pitch and dynamic level.

The way this 'Contoured Sound Modelling' works goes something as follows.

Let's suppose you've got a friendly grand piano to hand that doesn't mind sharing its timbral information provided you promise to polish it next week. If you play a note as loud as the piano is capable of, and sample it for as long as it takes to decay, you'll have gained a single timbral point from the 250 sounds that Kurzweil claims for each key. At this stage, if a computer was asked to analyse the sound and suggest how it would differ if played at half the intensity, its only clue would be the height of the constituent harmonics, so a key played with half the intensity would imply half the levels of all the harmonics over the entire course of the note.

To help the computer guess more accurately the next stage is actually to sample the note played at half intensity. This gives the computer a better representation of how the constituent harmonics change with intensity, so that if you then play a key at somewhere in between these two sampled intensities, the computer can interpolate between the two lots of sample information and come up with an educated guess. And, of course, it doesn't take a vast leap of the imagination to see that the more samples you take per note for different intensities of playing, the better the computer will be at filling in the gaps.

So what's being created is a system that improves its performance each time you give it more information. That, in a nutshell, is what Artificial Intelligence is all about, and what makes the 250's sounds tick. And according to Kurzweil's publicity info, the sound complexity captured by the 250 is 'at least 200 times greater than that of most other digital keyboard instruments'. Not 125,000 exactly, but still a lot closer to home.

Close Up

Make no mistake, the Kurzweil 250 is big. Big, black, and beautiful. Well, the last-mentioned lies in the eyes of the beholder, but remember that this near-as-dammit hundredweight of solidly constructed bulk won't spare those hernial orifices if it's manhandled solo. A bit like taking on Grace Jones without the assistance of Arnold Schwarzenegger... But there's more to come - the 22lb 'power pod' that doubles as power supply and foot pedal unit, to be exact. Clever idea, that - massive transformers are probably about as good as anything for preventing footpedal boxes from skating about on a polished floor. Also, putting the power supply on the floor gets around hum and heat problems.

The keyboard itself is an honest-to-goodness Pratt & Reed type covering an 88-note range. The keys are somewhat lighter in feel than the average concert grand, and they're also the standard synth length of 5.5 inches from back to front, but the keyboard plays like a very responsive grand, not a synth.

"Ever eager to take up the gauntlet, I dug out my dusty volumes of Rachmaninoff and Debussy, rang up a friend with a 9ft Bosendorfer concert grand, and arranged to drop by with the 250."

Central position of the control panel is occupied by a back-lit LCD. Underneath it, there's a trio of momentary contact switches (Yes, No, and Select) and a keypad with a cursor cluster to the right. Arranged around this focal centre are the Effect, Assignment, Program and Media control sections. To the keypad's immediate left lie buttons for Chorus, Split Keyboard, and Transpose, and next come switches and sliders for detuning and assigning functions to the two performance wheels. Finally, at the far left, we find the master tuning and balance/output level controls. On the keypad's right, you've got the section for program control - the means of setting up keyboards, instruments, and sequences - and at the far right, the media section for organising the 250's intercourse(!) with MIDI keyboards, sync receivers and senders, and computers. Oh, and before I forget, there's also a socket underneath the front of the keyboard ready to receive plug-in ROM sound cartridges (at £600 each), though there are conflicting reports as to if and when these will actually appear.

Unlike similar instruments that oblige you to wrestle with umpteen screws and hinged keyboards before you can find out what's where inside the box, the 250 adopts the blindingly sensible approach of putting all the circuitry on a sliding plate that can be pulled out from the back of the keyboard by merely undoing a trio of screws. Superb for servicing or replacing ROM boards. This manoeuvre reveals three large PCBs that normally stretch their way underneath the keyboard. The first of these gives pride of place to a 68000 processor plus all its attendant circuitry, together with 108K of battery backed-up CMOS RAM (16 x 6264s - a cool £500-worth on the UK market!). Next, the middle board, which houses the massive amount of memory embodied in 60 x 23256 ROMs - a total of 2Mbyte. Now, it's only recently that these 256K ROMs (32K x eight-bit) have appeared on the market, which is probably one of the reasons for the delay in the 250 coming off the production line. And at a one-off price of $25 for a 23256 ROM, it's not hard to see where a healthy slice of that £11,000 is going. However, unlike the average drum machine ROM, where the data in the chips is essentially a brute-force digital recording of the original, what's ensconced in the 250's ROMs is a load of tables that represent Kurzweil's AI-derived model of the sound. So, if a note's played on the keyboard, the 68000 trots off to the ROMs to yank out the necessary information on that sound played at that pitch and velocity, performs some impressive number crunching to reconstitute all the necessary sound data, and then sends it off to the third board to 12 AD7545 12-bit DACs, attendant sample & holds, and some well-shielded switched-capacitor type low-pass filters to create the instrument's 12 output channels.

Finally à derrière, there's a host of inputs and outputs to cover most contingencies. Specifically, and from right to left, the 250 provides Line In, Mic In, a couple of jacks for footpedals, headphones, balanced left and right outputs, low and high outputs, the main multicore power/footpedal socket, MIDI Out, Thru and In, Sync Out and In, Click Out, Trigger In, and last but far from least, a 37-way D-connector marked 'computer', more of which anon.

Even given this generous list of ins and outs, what stands out as a curious omission is separate outputs for each of the 12 channels. No doubt Kurzweil will argue that separate outputs are irrelevant on a performance keyboard like the 250, but given that it's possible to set up 13 different drum sounds on the keyboard (as Kurzweil have done themselves - not once, but four times - in the factory keyboard setups) and record each one separately into the multitrack sequencer, it seems strange not to put the final layer of icing on the cake and provide individual outs for the performance and (especially) recording situations that demand them. Apparently though, Kurzweil are giving this one some careful thought...

Sound Comparisons

Kurzweil have made a great deal of advertising mileage out of their claim that the 250 'perfectly recreates the sound quality, as well as the dynamic range, of any acoustic instrument - even a concert grand'. At this year's NAMM show, an A-B comparison was staged between a 'top-of-the-line' 9ft concert grand and the 250, both played through a $40,000 sound system. According to Kurzweil, the general consensus was that it was impossible to distinguish any difference between the two. Hence the motto adopted by the company: 'you can't tell the difference'. Well, ever eager to take up the gauntlet, I dug out my dusty volumes of Rachmaninoff and Debussy, rang up a friend with a 9ft Bosendorfer concert grand, and arranged to drop round with the 250. Much back-straining and estate car groaning later (the things I do for E&MM!), I put the comparison to the test, as well as going through the other factory sounds in the 250's ROM catalogue.

As far as the bass end of the piano was concerned, the poor old Bosendorfer hardly got a look in - even with its extra notes flap lifted. And yes, as some other reviewers have commented, the Kurzweil's piano does sound as if it's being piped in from a studio - albeit via some expensive monitors - rather than being actually there in front of you. In fact, it was a rather curious feeling - a sort of 'Out of Body' pianistic experience. But hang about, didn't I hear a difference in tone going from this note to that even though I was using the same touch? And what about that group of notes further up - isn't the hammer sound a little too strong for comfort? To be precise, what actually transpired from this critical going-over is that the 250's 88 notes split into about 16 clearly definable sampling regions, where both the notes below and notes above are timbrally different to their neighbours. The following notes illustrate what I noticed going up the grand piano keyboard from middle C:

C-E: bright sound, velocity leads to brighter tone
F-C: darker tone, velocity leads to more hammer sound; pitch of hammer increases going up the group
C-F: thinner, noticeable delay glitch with pitch wobble over entire group
G-F: close-miked sound, very prominent hammer tone
F-D: wooden, very pronounced hammer strike which changes pitch with the note
E-F: brighter and louder, less hammer
G-C: same brightness, hammer rises in pitch

If there was a general impression to be gained from this devious detective work, it was that the bottom note of each group was darker and had a lower-pitched hammer tone than the top. Superficially, this sounds pretty much like what we've come to expect from the multisampling technique - not enough samples to cope with going smoothly from one section of the keyboard to another. But that's only one side of the story. What about the effect of velocity on the timbre? Well, it's here that the 250 really shines: the way the tone darkens and lightens with dynamics is nothing short of stunning. During the comparisons, this prompted me to start burying myself in sheet music to see how the 250 could cope with the tonal ebb and flow of Rachmaninoff and the atmospheric haze of Debussy. Perhaps not surprisingly, the Bosendorfer won on both counts, but isn't that what you'd expect from a £25,000 concert grand? The fact that the Kurzweil was a dam good loser in this (extremely unfair) competition only goes to show that its designers have got a lot right. And when you remember that this particular 88-note piano has been squashed into just 512K of ROM with an utterly natural dynamic performance complete with the requisite timbral changes, a smooth glitch-free decay that stretches off into the distance for tens of seconds, and a 50kHz sampling rate to boot, it's difficult not to be a little awestruck by what Kurzweil have achieved.

The rest of the stock factory sounds are roughly what you'd expect from a sampling system that's aiming to prove its point to as wide a range of musicians as possible. Namely, a mixture of organs of all shapes and sizes with key clicks that turn into noise chiffs down at the bottom end, various string sections seemingly caught in a time-warp, sundry brass instruments of which only the baritone horn really stands out as being exceptional, the very quirky 'endless glissando', which'd put even a barber's pole in a twist, a superb acoustic bass (similar to the DX7's, but without the gut-wrenching strain that comes from digging into the keys), an acoustic guitar that isn't quite there and shows an alarming inconsistency of tone below middle C, and about 13 different drum sounds arranged in a variety of somewhat bizarre drum kits.

As in the case of the grand piano, all these instruments (with the exception of the velocity-less electric organs), prove the usefulness of Contoured Sound Modelling. Strings bite more deeply, horns get brighter, kick drums get thuddier. In short, the technique works triumphantly.

But at the same time, the problems that cropped up with the grand piano appear just as significantly on these sounds. Taking the fast-attack strings as an example, we find a rich 'cello at the bottom - real C-string stuff, this - that wends its way up the keyboard changing quality here and there. Perhaps that's to be expected when you've a 'cello turning into a viola into a violin, but that doesn't explain why a sustained note should give a very good impression of glitched looping. This was especially bad above the C above middle C, where the cycling background noise (the bow sound?) plus an all too obvious loop point give a good impression of an altogether too groovy string section. Going up, the D two octaves above middle C has an annoying chirp at the start of it, and then in the top fifth of the keyboard, all the notes have a twenties-like portamento at the start of them, effectively transforming what should be sweet-sounding fiddle lines into something reminiscent of the shower scene from Psycho...

The lack of quality control between the different factory sounds was something that struck me repeatedly during the time I had with the 250. Take the baritone horn, for instance - a rich, silky sound of character and quality, and an asset to any musician. Next to that in the brass line-up, we find a rather weak trumpet, whose over-emphasised 'wah' results in a sound that's more reminiscent of a harmonium than any member of the brass family, and the distinctly flabby trombone, with a pitch wobble that can only be described as flatulent. Move to the percussion and you'll find a host of powerful timps, toms, and kicks, snappy snares, and crashing cymbals (with a decent if not generous decay time of 2 seconds). In short, excellent percussion that responds as dynamically as all the other 250 sounds. But doesn't it seem a bit bizarre putting drum kit percussion on a performance keyboard? Sure, it's a good way of extending the number of sounds on the ROM board up to the quoted 30, but shouldn't the remaining ten keyboard-orientated sounds be uniformly better than they are in reality?

"Of the currently available options, perhaps the most interesting is to customise a particular instrument with its own ambient surroundings, courtesy of the Kurzweil's chorus system."

Modifications

Kurzweil's original promotional material made great play of the modifications that can be applied to sounds, both sampled and preset. The original intention of offering programmable VCFs seems to have been ditched in the production version, but what is now offered - 20 different functions for echo/chorusing, vibrato, tremolo, pitch bend, and envelope control - should be enough for most aspiring Kurzweilers.

And if it isn't, there's always the Sound Laboratory software to come, which should allow you and an Apple Macintosh computer to indulge in the analysis of sounds, simulation of filter sweeps, mergings, cross-fades, or whatever. In short, whatever your favourite form of sound manipulation may be, the Sound Laboratory should provide the means for its realisation.

Of the currently available options, perhaps the most interesting is to customise a particular instrument with its own ambient surroundings courtesy of the Kurzweil's chorus system. The main problem with this is that all the delay line effects consume extra channels, and the reason for this is that the effects are achieved by subtly-applied pitch offsets and staggered envelopes rather than a true recycling of a sound through RAM. Again, that's one of the limitations imposed on the 250 by ROM-based sound storage.

There are four types of effect available: doubling, full chorus, flanging, and echo. Full chorus is the greediest of the four in that it spawns chorused voices on either side of the root voice, thereby reducing the keyboard's 12-note polyphony to just four-note. In practice, it's not that bad because of the dynamic channel assignment, but you've still got to watch your Ps and Qs pretty closely. Personally, I'd be more inclined to add on a MIDI-controllable DDL (like the Yamaha D1500 or Powertran MCS1) than mess around with predefining delay line effects.

A further modification (of a sort) is the facility to combine different sounds across the keyboard, both as conventional splits and specially layered set-ups. Indeed, of the 40 keyboard set-ups preset in the 250, only 13 are of single instruments, the remainder including a wide variety of split arrangements (ranging from an acoustic bass/piano combination, with a single split at B2, to multiple-split drum kits) and more or less complex layerings (the slow strings and guitar or octave-detuned layered guitars, for instance). In fact, each keyboard set-up can have as many as six layers. Within each single instrument layer, the sound can be transposed, pitch-shifted, timbre-shifted (within the limits of LFO modulation and delay line tactics), or velocity-adjusted (less or more sensitive, or even anti-velocity sensing), and up to 50 of these user-defined set-ups can then be saved in the backed-up RAM. The advantage of all this is that particular keyboard set-ups will then function as automatic arranging tools, playing several instruments from a single key - if that's what turns you on. The disadvantage is that, like the 250's chorus system, layering uses up yet more channels.

Sequencing

Again, the sequencer looks wonderful on paper.

It would take a pedant and a half to complain about specs that run to 12 independent realtime polyphonic tracks, 8500-note battery-backed RAM (with the expansion RAM in place, which, depending on whom you ask or wave a bunch of green ones at, may or may not be fitted as standard), all manner of sequence and track editing, playback quantisation, multi-input tempo control (whatever that may be), and variable-rate external sync. However, the majority of these functions were missing from the software in the 250 I was using. Indeed, as you read through the sequencer section in the beautifully-Macintoshed manual, you come across comments such as 'you cannot currently edit sequences', 'none of the special effects (including sustain) are recorded in the sequence track (this will be remedied in future releases)', 'as yet quantization is not available', 'tempo tap is not yet implemented', and 'this version of the sequencer is not doing append correctly'. Nuff said.

Don't get me wrong. The 250's sequencer isn't a bad sequencer, it's just an unfinished one. Which raises the question as to whether the release of this version was perhaps just a little too premature for Kurzweil's own good, bearing in mind how difficult it is to review features that are outside current reality. Another problem is that as the 250 stands, there's no way in which either instrument parameters, keyboard set-ups or sequences can actually be saved external to the battery backed-up RAM, aside from putting a £2000 Macintosh on-line to the 250. Again, this reflects an about-turn from Kurzweil's original promo material, in which an artist's impression of the control panel showed cassette controls and the specs list included 'off-line digital cassette storage'.

What I can say is that the sequencer works very well as far as it goes. But you can understand the frustration I felt when, having played in a track using the grand piano keyboard set-up, I discovered that the sequencer had omitted to lodge the function of the sustain pedal in its frame of consciousness. True, the manual informs you as such, but that doesn't stop it being a hell of a let-down after all the hype and expectation behind the 250. Still, the multitracking works more or less painlessly and the LCD is used to good effect to keep you informed about where your tracks and sequences are going, a factor that's vital if you want to avoid inadvertently erasing a sequence.

One point that's vitally important to bear in mind when using the sequencer is the fact that the 250 has 12 output channels. So, if you go overboard with your polyphony on the 12 sequencer tracks, don't be surprised if you fail to hear everything you put in. Remember that the thing's only human, and you'll be all right. In fact the 250's software includes a variety of different 'channel stealing algorithms' (also a feature of the Synergy, as it happens) designed to reallocate channels dynamically as and when they're needed, without giving the idea that notes have suddenly been cut off in their prime. I was impressed by the operation of this: even with lots of sustain and big power chords, you got the impression there were more than just 12 channels at work. Mind you, with multitrack facilities to hand, it would obviously be a safer proposition to use the 250's tape sync (when it's working, of course) to stagger recording the 12 potential sequencer tracks onto tape.

Anyhow, having laid down a couple of ditties on top of an excellent, driving drum pattern some kind personage had left in the sequencer, I scanned through the remainder of the 25 sequences that can be stored onboard just to see what others had been up to. And lo and behold, what should I come across but an 'arrangement' of Handel's Arrival of the Queen of Sheba, scored for fast strings and baritone horn presets. This proved a revelation, though not quite in the sense you'd expect. True, the intended string orchestra sounded as if it was playing most of the notes, but the quality of the playing and sound was such that it was all too reminiscent of those awful 78 recordings of ad hoc orchestras waging war with the classics. Mind you, not all the blame can be laid at the feet of the strings samples, because the lack of sustain was yet another thorn in the flesh. Which left the two baritone horns truly out-front with a style of playing and quality of sound that was completely at odds with the backing - though stunning nonetheless. Somehow, I don't think even the Alan Parsons Project would put up with these strings...

Future Prospects

The future of the 250 is really summed up by that connector on the back of the keyboard marked 'computer'. Kurzweil ditched their efforts in the direction of linking the 250 with the IBM PC some time ago which, bearing in mind the comparative slowness of the IBM, is hardly something worth losing much sleep over. More to the point is the fact that the IBM PC's 8088 processor only has an eight-bit data bus, whereas Kurzweil's current chosen computer, the Apple Macintosh, uses the same 68000 processor as the 250, which obviously makes for easier communication and software compatibility.

The first, and for 99% of musicians, most crucial expansion facility is sampling. As things stand, the 250 can't sample unless some additional hardware and software is pressed into service. This comes under the name of the Sound Digitizer, and comprises a set of PCBs, the first holding 512K of conventional RAM (no battery back-up) and plugging into the 68000 board, the second including an anti-aliasing filter, 12-bit ADC, and sampling software in ROM and plugging into the analogue board. This combination is expected to retail for between £1000 and £1500, and should make its appearance early in 1985. What you'll then be able to do is sample up to 20 seconds of sound into the extra RAM (at a sampling rate contingent on sound length and bandwidth requirements), perform a limited degree of analysis on it (along the lines of Contoured Sound Modelling, but without all the bells and whistles), and then use the sound(s) in conjunction with the ROM board presets. However, this welcome addition carries the major penalty point that you won't be able to save your own samples unless you also attach at £2000 Apple Macintosh computer.

Exactly what the Composition/Notation software package will offer isn't clear at present, but the main aims would seem to be a non realtime transcribing facility that puts any piece of music played on the 250 onto the screen of an Apple Mac, a printout feature that supports a gamut of possibilities from orchestral to vocal score preparation, and perhaps most importantly, the implementation of 'KMS Language', Kurzweil's own MCL.

One major problem is going to be the cost of all the software enhancements necessary for getting a Mac communicating with a 250: namely £2500 for the Sound Modelling Program, £1670 for the Sound Laboratory, and a further £1670 for the Composition/Notation Software. Add on the approximate figure of £1200 for the Sound Digitizer, and that's £7040 on top of the £10,995, making the princely sum of £18,035 in all: not a figure to be sneezed at. And that's still without the extra £850 quoted for each alternative set of ROM-based sounds.

Finally, there's the MIDI side of the 250. The fact that Kurzweil have devoted 34 pages of the manual to this feature is a clear sign of their intent. More importantly for the uninitiated user, the manual takes great pains to guide the average MIDI interfacer over the bumpy terrain of getting one instrument to communicate with another. In fact, the 250's features vis-a-vis MIDI are pretty comprehensive, and include programmable channel transmit and receive and the capacity to utilise system exclusive information.

Conclusions

The fact that the 250 has so much expensive ROM inside it highlights the major problems associated with ROM-based sounds. If all 30 sounds were of equal value, that wouldn't be so bad, but the truth is that some of them are really rather poor, so an upgrade has got to be on the cards (pun intended) in the near future. The alternative is to make sure that when you buy your 250, you have a good listen to the different ROM sound boards that are available. After all, if you're not a piano man (or woman - Production Ed) it's hardly worth sacrificing a quarter of the ROM space just for that one sound.

Doing the Devil's advocate bit, I have to say that I'm not convinced by the usefulness of ROM sound storage in a machine of this calibre and cost. It's fine for the drum machine market, where sounds are by tradition 'preset', and relatively high volume sales are the norm, but for this sort of quality keyboard, I'd have put my money on replacing the 2Mbytes of ROM with an equivalent amount of (cheaper) RAM, and then adding on a nice, fast (using DMA) double or quad density 5.25" disk drive for loading and saving sounds to and from RAM as and when you sample or need them. Indeed, it seems perverse in the extreme to go in the opposite direction to E-mu Systems, who've just added a further disk drive to the Emulator II (see review, E&MM November) in order to improve storage facilities.

Still, dress the 250 up with a nice, clean Macintosh and you've got a dual-68000 system that outstrips all its competitors in the areas of sound quality, hardware design, and software intelligence. So I back the principles, if not the practicality, of the 250 wholeheartedly.

I can't help thinking that what I've been looking at represents the tip of the 250 iceberg, given the promises its designers are making for the future. And of course, there's still the question as to whether Kurzweil have any intentions to licence their software techniques to other manufacturers. That would really put the Artificially Intelligent cat amongst the eight-bit pigeons.

Further information on all things Kurzweil can be had from Syco, (Contact Details); or Scenic Sounds Equipment, (Contact Details). Review model supplied by Scenic Sounds.