Vol. 30 Issue 3 Reviews

Composers’ Desktop Project version 5.0.1

Composers’ Desktop Project 5.0.1, from US$ 300 (various options are available, depnding on platform, GUI choices, and site licensing). Contact Composers' Desktop Project, 12 Goodwood Way, Cepen Park South, Chippenham, Wiltshire SN14 0SY, UK; telephone (+44) 124-946-1361; electronic mail archer@trans4um.demon.co.uk; Web www.bath.ac.uk/~masjpf/CDP/CDP.htm.

Reviewed by Jon Forshee
Paris, France

We make only the music we are able to make. As we choose our music-making environments it is often the case that we seek an environment that imposes the least restrictions upon our music-making activity; or, expressed another way, we may seek an environment that enables us as agents of our self-determined compositional guidelines.

The release of version 5.0.1 of Composers' Desktop Project (CDP) realizes the latest installment of just such a music-making environment. Initiated in 1986, the Composers' Desktop Project is a cooperatively developed software system, based in the UK, "designed specifically to transform existing sound samples for musical purposes, mostly via off-line processing." Sound sample transformation is achieved through a broad diversity of more than 500 command-line programs, and CDP includes two Graphic User Interfaces: Sound Loom, and Soundshaper.

Version 5.0.1 provides powerful additions to CDP's gamut of functions and utilities, including more than 40 new programs for soundfile transformation and speech processing, support for the 24-bit/96 kHz sampling rate (DVD quality), expanded documentation, and revision of the iconic Sound Loom interface. Notably (and finally!), CDP and Sound Loom now exist in a port for Mac OS X. In this review I demonstrate some of the advantages of this new version from an actual work session; I shall also mention a few peculiarities unique to CDP. I am assuming the reader has some familiarity with CDP and the Sound Loom interface, as a detailed discussion of each procedure used would be prohibitive in the present context.

I am using the new Mac OS X port of Composers' Desktop Project, with the Sound Loom interface, version 9.2b, with the new color scheme by Dale Perkins of Leeds University. [Note: during the writing of this review Sound Loom version 9.7.3 was released by Trevor Wishart, and is available from his Web site (www.trevorwishart.co.uk/).] My current musical work concerns an electroacoustic composition, in four channels, which uses recordings of readings by poet Charles Stein as source material. I have a general outline, sketched beforehand, of the musical terrain of this composition, including some of the ways in which the piece will highlight unique properties of Mr. Stein's voice.

The Soundfile Editor
My source samples exist as large files containing entire readings; I intend to work with smaller portions of the sources, so I select one of the recordings and set the target directory containing the file in Sound Loom. Once Sound Loom has loaded the directory, I “grab” the file and add it to the workspace portion of the interface. When the file is listed in the workspace window, I invoke Sound Loom's new Soundfile Editor by ?-clicking on the file. Sound Loom's Soundfile Editor offers all the basic functions of a graphic soundfile editor, plus features for superimposing and editing amplitude and pan breakpoints. I quickly locate the passage I want to trim and select the area by shift-selecting the region with the mouse. Selecting regions draws a solid block around the region, obscuring the graphic representation of the soundfile behind it. Saving the trim involves giving the file a new name and pressing “Savesnd.”

Throughout this process some differences between the Windows- and Macintosh-platform versions of Sound Loom become evident. For example, the layout of the Soundfile Editor is the same, but to accommodate the familiar feel and style of the Mac OS, the button text is abbreviated: where the PC Sound Loom reads “Save Sound,” the text in the Mac-platform Sound Loom is abbreviated to “Savesnd.” Similar abbreviations appear throughout the Mac interface, and for those who are familiar with Sound Loom on the PC it may take a moment to re-orient to the Macintosh version, even though the layout is identical in all other respects.
Before I can proceed, my trimmed sample should be normalized, accomplished quickly enough by entering “Chosen Files” mode in the Sound Loom Workspace, moving the file I created to the “Chosen Files” window, hitting “Process,” choosing the group LOUDNESS, and selecting Normalize from the menu. I save the file, return to the workspace, and am now ready to audition the sample.

Speech Processing
One of Mr. Stein's unique vocal abilities is what he refers to as “Sound-Text Poetry,” which, at first hearing, sounds a little like speaking in tongues, or riffing in what may sound like Italian or Hindi. My first sample is a terrific instance of “sound-text-speech” in “Italian,” and features the poet peppering his speech with rolled r’s. I intend to isolate Mr. Stein's r's, and dramatically extend the rolling while preserving the flow of his sound-text speech.

Among the many new features of version 5.0.1 are a host of programs geared toward speech processing; one of these, R_EXTEND, in the GRAIN group, is designed for extending iterative sounds. Specifically, R_EXTEND can pick out iterative sounds in a sample and extend the iteration. The iteration, in this case Mr. Stein's rolled r's, can either be slightly extended within the sample, thereby drawing more attention to them, or the rolling can be extended beyond recognizability, perhaps creating an occasion for morphing with another sample later on.

The R_EXTEND function is invoked in the GRAIN group, and comes in two flavors, one for manually specifying the location of iteration, and one which finds iterative locations based upon user-adjusted parameters. I start with the first flavor since I have a particular sound in mind, and do not wish to be led too far afield by interesting accidents just yet.

As with all the programs included in CDP, the parameters for R_EXTEND are comprehensive, and may not seem intuitive at first; however, what the parameters do becomes evident with use, and I find that a little fidgeting with parameters I don't understand is revealing. To simplify my first attempt, I re-trim the soundfile to include just one instance of rolled r's. I set, by estimate, the start and end of the iterative section with the first two parameters, and I set the length of time-stretching (that is, the “extending”), in the third parameter. Additionally, I set the anticipated number of iterative segments, the range, in octaves, of the segments to be found, as well as the maximum number of adjacent occurences (overlap?) of any segment. In a spirit of play I adjust, but don't think too much about, the last two parameters, amplitude scatter and pitch scatter. The final option in the list offers the possibility of keeping the extended iteration only. I opt to keep the extended iteration within the flow of Mr. Stein's speech.

I execute the function and, as I expected, am a little off in my guesswork. The extended iteration and timestretching occur just a little after the actual rolled r's in the sample, so I return to the parameters page and, after a few more attempts, nail the rolled r's, producing the desired meta-linguistic speech sequence. During the guesswork, if I enter a value for a parameter that contradicts another parameter's setting, I will receive an error message in the Run window, telling me which parameter needs adjusting.

A useful function in conjunction with R_EXTEND is another new addition to the GRAIN group, ASSESS, or as it appears in the menu, “assess max. number of grains.” This handy utility scans the soundfile and reports amplitude and window size that will result in the greatest number of grains. This is a useful feature when using any of CDP's granular synthesis functions, and I also use this utility with good results when searching soundfiles for suggestive regions amenable to morphing.

After experimenting back and forth with these two utilities I am able to achieve one segment of a larger gesture which, overall, was suggested by the content of the sample and which fit in with my pre-compositional sketches. There is more work to be done with these new functions, yet my personal work habits engender a degree of play, so I proceed with shorter experiments with a few other additions to this version of CDP.

SHUDDER applies a time-variable amplitude modulation to a soundfile,
creating a "shaking" effect. I found good use of this utility by sculpting an
accelerando in the modulation which "leads" to an outburst of thickly layered
timbral polyphony, a section which includes several layers of Mr. Stein's speech

Similarly, SPACEFORM provides parameters for a “sinusoidal distribution” within the stereo field, in which a soundfile or portion of a soundfile may be swung between two channels. A first run of this function reminds me of a “spatial” form of amplitude modulation, so I engineer an accelerando between left and right channels similar to what I achieved with SHUDDER, and this particular realization is put aside for later use.

ENVEL TIMEGRID parses a soundfile into sections separated by slices of silence; the space between chunks is variable, as are chunk lengths. An immediate application in the present piece might involve cutting a long-ish phrase by Mr. Stein into 8, 12, or 16 parts, each of which is later assigned to one of the four channels, possibly affording the perception of a single, connected utterance dispersed throughout the quadraphonic space.

I have used CDP throughout as an environment in which to reify my imagined musical profiles, and I have had success with the functions I chose; moreover, in my trial and error approach to these new functions I have encountered possibilities I did not envision, which may yet prove fertile ground for further development. There are many more new functions in version 5.0.1 than I have discussed. In particular, the new joining and sequencing programs, JOINSEQ, JOINDYN, SEQUENCY, and SEQUENCE2 deserve attention for those who have used CDP's non-track-based assembly routines.

With respect to the functions I did use, a little poking around was in order to achieve what I wanted, and despite the revisions (and port to Mac OS X), the Sound Loom interface remains a little obtuse at first use. Additionally, it is apparent that with CDP and Sound Loom I work best on small sections, creating for me a micro-to-macro method of developing the material. So much is this the case that I choose to sequence sculpted passages created in CDP in a third-party track-based software sequencer, as opposed to CDP's in-house sound object assembly sequencer. This is only a personal preference, however, and one needs only to listen to recent compositions of Trevor Wishart to hear just how flexible CDP's non-track-based sequencer is.

Through the cooperative effort of Composers' Desktop Project, CDP and Sound Loom are developing quickly; a glance at the “Potential Project Page” of the CDP Web site reveals enticing possibilities for the near future. However, the Mac OS X port still takes a little effort to install. In order to successfully install Composers' Desktop Project and Sound Loom on a Macintosh computer, it is vital to read Mr. Wishart's installation instructions (found at www.trevorwishart.co.uk/slfull.html).

As the present context does not allow for a full anatomization of CDP's many new (and upcoming) features, visit the Composers' Desktop Project home page (www.bath.ac.uk/~masjpf/CDP/index.html), where much more is offered, including new tutorials and, most important,  information on how to purchase the system.