Prosoniq, the German software company behind many of the DSP processing algorithms used in MIDI + Audio sequencers, have now made the powerful fruits of their expertise available in stand-alone form. But can the software actually be put to good use? PAUL D LEHRMAN investigates.
SonicWORX Artist is fascinating audio editing software for the Macintosh from a German developer called Prosoniq. Though their name may seem unfamiliar, you may have seen or used the products of their labours, as Emagic have licensed Prosoniq's audio processing algorithms for use in Logic Audio (the pitch-shifting algorithm found in Logic's Time Machine II, for example, is a Prosoniq development, as is the on-board EQ, flanger, and reverb), and the company also produces third-party plug-ins for Cubase VST.
pros & cons
PROSONIQ SONICWORX ARTIST v1.0.2 £299
Tons of algorithms.
Good sound quality (when you want it).
Can make a sound file out of anything.
Clumsy user interface.
Obscure, unexplained processing parameters.
Often impossible to make it do what you want.
SonicWORX Artist contains some of the most powerful sound-modifying tools we've seen yet, and the program is potentially brilliant, but so far it is critically hampered by an uninformative manual and obscure operating parameters.
The program is one of three from the company: another package, SonicWORX Studio, is designed for mastering and post-production facilities, and emphasises practical functions like de-clicking, phase alignment -- of concern to those still cutting vinyl -- and sample-rate conversion, rather than the more creative functions found in the Artist package. The third product, Professional, promised for later in 1998, will combine the two existing products, and add a voice-extraction feature which can either isolate or eliminate lead vocal tracks from a mix.
Installation is from three floppy disks, and couldn't be simpler, with copy protection provided by a hardware dongle. The software uses the internal Macintosh sound drivers (with Sound Manager 3.1 or later), and will work with Digidesign I/O hardware provided you have the proper drivers and INITs in your System folder.
SonicWORX Artist was originally conceived for Silicon Graphics workstations, but the port to the Macintosh is pretty seamless. Theoretically, this version of the program will work on both 680x0- and PowerPC-based computers, but on older machines I imagine the wait times for the processing would be unbearably long. In fact, the manual mentions how the 'save after execute' function (which is useful for processing which takes overnight to finish) will be 'appreciated' by 68k Mac owners. While overnight rendering is a fact of life in the computer graphics world, no sound designers I know have that kind of patience. The Studio version of the software won't run on 68k Macs at all [as this review was being prepared, Prosoniq announced early plans for version 1.5 of SonicWORX Artist, in which all support for 68k Macs has been abandoned -- Assistant Ed].
Since the program does all of its storage and housekeeping on hard disk, it runs surprisingly lean, requiring only 8Mb of RAM to function quite smoothly. Of course, you'll want plenty of empty disk space to store your experiments, but you know that.
On the face of it, the user interface is very simple. There are two windows: one to display the waveform, and the other to manipulate the processing parameters. When a file is open in the waveform window, you can cut and paste or otherwise move it around, and select part or all of it for processing. You can specify different parts, or 'ranges', of the file as input and output 'channels' -- for example, take a processed version of the first part of a sample and replace the second part of the sample with it; or if you are using an algorithm with two inputs, such as Morphing, you select two input ranges and a third range for output. You can have multiple files open, and they can interact with each other: a range from one can be used as an input channel to affect the audio in another.
You can open any standard audio file format, although the manual recommends converting WAV files into AIFF format before you open them, using some other program (would it really have been so hard to build this in?), because of the way offsets are handled in the WAV format. In fact, as with the well-known audio editing shareware SoundHack, you can open any binary file, from text, to pictures, to an Excel spreadsheet, in the program and treat it as audio data. No doubt this has some fun applications, but unfortunately neither the manual nor the example files that come with the program tell you anything about the creative possibilities this could afford you.
When it comes to saving files, the program will create mono or stereo files, using AIFF or Sound Designer II format, with 8-, 16-, or 24-bit word lengths, at any sample rate. There are no provisions for exporting to a sampler over MIDI or SMDI, or to a Digidesign Samplecell card -- you'll need Peak or Samplecell Editor for that.
The second window is where you set up your processing parameters. You open a new one of these with the New Parameter File function under the File menu, and select the algorithm you want to start with from the Algorithms menu on the left of the window. The various algorithms are provided as plug-ins, similar to (but not compatible with) Digidesign and Adobe Premier plug-ins, which go into a folder on your Mac alongside the application itself. The version of SonicWORX Artist I reviewed came with 57 algorithms, but an ever-growing selection of additional plug-ins, including some designed by users, is available from Prosoniq's website (www.prosoniq.com). The plug-ins themselves are very small -- anywhere from about 9K to about 70K in size -- so downloading groups of them takes very little time.
You can have multiple algorithms working at the same time, and you can save the lot as a Parameter File, which recalls all the algorithms in the order you loaded them, plus all of their parameters. This is a very important feature, as trying to recall which algorithms and which parameters do what can be very confusing, as we shall see.
The plug-ins are organised into 10 groups, each with its own sub-menu, and offer an astonishing variety of processing options. One group (the so-called 'Utilities') consists of fairly conventional algorithms, such as Fade, DC offset removal, Normalise, Reverse, Pitch-shift and sample-rate conversion. With the group entitled 'Mathematic Operations' you can combine files, subtract them from each other, multiply them for interesting modulation effects, and do Emu-style transform multiplication, which combines the spectra of two different sounds to create something with the characteristics of both.
The 'Timebase and Pitch conversion' group provides the usual pitch-shift and time-compression/expansion functions, but it also has a Pitch Designer, which breaks the sound down into various tonal components and changes their pitch independently -- so for example, you can literally isolate a kick drum and raise it up in pitch without affecting the rest of the track. The Dynamic Evolution parameter even changes the amount of pitch change over time. In the 'Special EFX/Sound Design' category are telephone filters, vinyl-noise creators, drum loop "jungle-isers", an algorithm that uses pitch-shifting and delays to create 'crowds' from a recording of a single voice, and others that add various types of unusual distortion, including a wild so-called Wavelet Alienise feature, which creates harmonic distortions that are pitched inversely to the original source.
Finally, there are the 'Audio Rendering' plug-ins, which include several different morphing algorithms, and an Apply Features function that imposes the timbral features of one sound on another.
Once you set up your parameter file (and you can have multiple parameter files open on the screen, although only one can be active at a time), you select the waveform you want to process, and click the big Execute button in the waveform window. The button gives you a progress report on the processing (in percent), and the waveform is redrawn at the end. Operation seems generally quite fast. The speed of any processing, obviously, will depend on the length of the file and the complexity of the algorithm(s), not to mention the speed of your computer, but on a 100MHz Power Mac, using mostly files a few seconds long, I never felt I had to wait unduly. An extremely useful feature (especially because there is no Preview function per se) lets you listen to the processed parts of the waveform before the software has finished calculating, and if you don't like what you hear, you can abort the operation by pressing the usual Command-full stop key combination.
If all this is starting to make you drool, I don't blame you. There's an enormous amount of power under the bonnet of this program, and the transmogrifications it can perform on your sounds are astounding. In just a few minutes of fooling around, I was able to create a singing cymbal, turn Dusty Springfield into Darth Vader (and still have her remain on-key, more or less), and make a drum loop sound like it was falling off a lorry in a tunnel full of water.
HOW DID I GET HERE?
The downside of all this, however, is that I couldn't possibly tell you what I did to create these effects. I could save the parameter files (and I did), but since what a sound processing algorithm does (in any application) is so dependent on the source material, were I to give you the parameter files for you to use on your sound files, you would get completely different results. I'm not exactly a novice at sound design, but I quite simply had no idea what I was doing when I was setting up the various algorithms. And this is the fault of Prosoniq, who have fallen down badly in explaining to the user what is going on in its product (see the 'Flying Blind' box elsewhere in this article). It's really a shame, because this could be a formidable tool, if only anyone outside of its designers could figure out how to use it.
The software itself doesn't help much, and there are some counter-intuitive features that will create even more confusion. As I mentioned earlier, there are no precise timing indicators anywhere, so setting up edits that follow exact timings or tempos is pretty difficult. The two-window construction means that you are constantly mousing back and forth to make adjustments, and to activate something in a non-active window, you have to double-click on it. You can avoid this by using command-click, but that ties up one hand.
A notable exception is the Execute button in the waveform window, which regardless of which window is currently active, requires only one click. Unfortunately, that button's visual response is delayed, so you may not think you have pressed it when you actually have; if an operation is very fast, you can end up executing it twice. There is only a single layer of Undo, so it's very easy to mess things up to the point where you can't get them back -- BIAS Peak's unlimited, catalogued Undo feature would have been extremely welcome here. There's also no Redo function, but there is at least a Compare feature, which lets you listen to the last version of a file without undoing the most recent bit of processing you've done. But once you Undo it, it's gone. Some attempt is made to explain what the program is doing at the beginning of the manual; there's a little essay about applications of neural nets in audio processing, and there's a short discussion buried in the section on the software's Preferences menu item, about the so-called 'Multiple Component Feature Extraction' process that it uses to analyse files. But both of these texts are woefully inadequate when it comes to explaining how to actually use the program. There is an introduction that contains a one-page tutorial which will only have you scratching your head, and there are no tutorial or even example files included with the software, except for four rather silly audio files. Perhaps the developers don't think users are really capable of understanding what's going on, so why should they bother writing an informative manual? Or perhaps they think all users are so knowledgable about these concepts that they don't need a manual -- the 'Real engineers don't read the docs' attitude. Perhaps they just didn't think it was worth spending the money to write a good manual, or perhaps they are being deliberately obscure in order to protect their processes from imitators. Or perhaps the whole thing is really just smoke and mirrors, and the concepts they explain so perfunctorily aren't really worth much at all. But whatever the reason (and you are welcome to think up your own), they do themselves a major disservice by keeping potential users so ignorant.
The manual, besides being not particularly well translated from the German, is vague and unhelpful about how the various processing algorithms on offer work, what they are doing, and what the parameters mean. The more complex the algorithm, the worse it gets -- many of the most obscure and important parameters are expressed in 'units' that have nothing to do with the real world, and so setting them is an exercise in flying blind.
Some attempt is made to explain what the program is doing at the beginning of the manual; there's a little essay about applications of neural nets in audio processing, and there's a short discussion buried in the section on the software's Preferences menu item, about the so-called 'Multiple Component Feature Extraction' process that it uses to analyse files. But both of these texts are woefully inadequate when it comes to explaining how to actually use the program. There is an introduction that contains a one-page tutorial which will only have you scratching your head, and there are no tutorial or even example files included with the software, except for four rather silly audio files.
Perhaps the developers don't think users are really capable of understanding what's going on, so why should they bother writing an informative manual? Or perhaps they think all users are so knowledgable about these concepts that they don't need a manual -- the 'Real engineers don't read the docs' attitude. Perhaps they just didn't think it was worth spending the money to write a good manual, or perhaps they are being deliberately obscure in order to protect their processes from imitators. Or perhaps the whole thing is really just smoke and mirrors, and the concepts they explain so perfunctorily aren't really worth much at all. But whatever the reason (and you are welcome to think up your own), they do themselves a major disservice by keeping potential users so ignorant.
Speaking of buttons, there's no on-screen Play button -- you have to use the space bar -- but since the biggest button on the screen is Execute, I found myself hitting that quite often when all I wanted to do was hear the file. Needless to say, this caused a certain degree of upset.
Other aspects of the user interface are also annoying. The parameter controls use 'fluid' bar graphs which give it a nice feel, but don't add much to the program. The input and output ranges in the waveform window look like the markers in any other sound-editing program, but you can't move them by hand: you can only select a region and say "Put 'em there!" If you want to move just one without disturbing the other, this can be a pain. You can easily select the area within a range, but there's no way to select quickly the area outside a range, so there's no fast and accurate way to, for example, break up a file into two ranges that butt up against each other. As you click around, it's remarkably easy to get the waveform window confused -- it will actually show parts of selected regions as un-selected, and vice versa. Much of the time, therefore, you have no idea what you're looking at, and have to re-size the window (there's a Redraw function, but it doesn't fix this) to get it right.
Once you bring more than one algorithm into a parameter window, you can't change the order in which they process the sound; any new algorithm brought in goes at the end of the list. Should you want to place it anywhere else -- for example, put the de-reverberator in front of a vocoder -- you have to delete all the algorithms that come after it, load the new algorithm, and then re-load the ones you just deleted. Unfortunately, all of the parameter settings for the algorithms you've deleted and re-loaded are then lost! It might help somewhat if you could use other parameter files for temporary storage of modules, but since you can't cut and paste algorithms between parameter files, that doesn't work. There is also no command-key equivalent for Delete in the parameters window, which means a lot of extra mousing.
The way the program handles stereo channels and inputs and outputs is confusing and inconsistent. I got it to create a 3-channel Sound Designer II file, which is something of a crime against nature -- no other program will deal with it. Some of the processing algorithms are designed to handle stereo audio, and provide proper correlation between the two channels when they are being processed, while others aren't -- but you have no way of knowing which is which.
And am I being too picky when I complain that some of the dialogue windows (like the Save As... one) have not been translated from the German? I know what a Cancel button looks like, but what about the poor mono-lingual English speaker who's never seen one before and is confronted with a button labelled 'Abbrechen'?
YOU CAN'T ALWAYS GET...
As for the quality of SonicWORX Artist's processing, it's a little hard to evaluate. The pitch-shifting and time-compression algorithms seem no better or worse than those found in other programs. The room simulator is decent, and the reverb remover is interesting, but unconvincing. The morphing algorithms are very powerful, and certainly do radical things to the sound. In general, it's extremely easy to make ugly sounds with a great many of the algorithms -- in fact, some of them (Spectral Erosion, for example) are expressly designed for that purpose. Certainly, if nasty is what you're after, you'll find plenty to play with here.
But making sounds that are interesting without being ugly takes a little more care, and given the vagaries of the user interface and the documentation, that can be difficult. Without a good understanding of what the algorithms are doing, the user is at their mercy, and (except for the most rudimentary processing) deliberately trying to go after something specific, as opposed to stumbling over something that you might be able to use, is pretty much out of the question. There's a lot to be said for experimenting, and fooling around, and arriving at things serendipitously. But for users to want to do that, using the software needs to be fun; and this program is not that.
In its favour, the software is quite robust, and I experienced few operational problems. I was, however, able to get it into an inescapable loop by trying to graft a sample recorded at one sampling rate onto the end of a sample at a different rate. I was presented with an error message, which I could click away, but then it would come right back. I had to force-quit the program, and of course I then lost all my unsaved work up to that point.
Working with SonicWORX Artist is sort of like driving a Lamborghini with one forward gear: you know there's a lot of horsepower there, and it may get you where you want to go eventually, but it's not very enjoyable. The developers are to be applauded for making so much power available to Macintosh users, but until they straighten out their user interface and documentation issues, I would look elsewhere for my sound-processing toys.
SonicWORX Artist £299; SonicWORX Studio £599. Prices include VAT.
Turnkey, 114-116 Charing Cross Road, London WC2H 0DT.
0171 379 5148.
0171 379 0093.