MASTERING META SYNTH - EMusician

MASTERING META SYNTH

MetaSynth 2.7 puts a painter's palette in your sonic toolbox.The multifaceted MetaSynth 2.7 sample editor and synthesizer from U&I software offers a broad
Author:
Publish date:

MetaSynth 2.7 puts a painter's palette in your sonic toolbox.

The multifaceted MetaSynth 2.7 sample editor and synthesizer from U&I software offers a broad range of DSP functions. But what really sets it apart is its palette of graphic sound-design tools. Not only can its Image Synth turn images into sound, but it also converts sound files to sonograms for graphic manipulation and resynthesis. MetaSynth's assortment of standard and not-so-standard DSP tools offers a range of possibilities not found anywhere else. This makes it an essential tool for both sound designers and composers.

Yet because MetaSynth is such a deep program and, frankly, doesn't conform to conventional Mac interface design, unlocking some of its mysteries requires a bit of know-how. In this article, I will provide an overview of some of MetaSynth's more advanced features, including the Image Synth, the extensive array of filters, and the Cross Convolution tool. Along the way, I'll touch on other aspects of the program, but keep in mind that this only scratches the surface of MetaSynth's capabilities.

IN THE HEART OF THE IMAGE SYNTHTo fully appreciate the uniqueness of MetaSynth's Image Synth, here's a brief look at how it works. At its heart, the Image Synth is a graphics editor with tools specially selected for sound manipulation. In converting a picture to sound, think of each horizontal line of pixels as representing a single note that will play the built-in synthesizer or a sample of your choice (the sound-making option is called the "rendering source"). The line's vertical position indicates the note's pitch, the horizontal position marks its position in time, and the line's length shows its duration. The line also provides time-varying volume and pan information for the note: changes in brightness control volume, and changes in color on the red/green scale control pan position. (The RGB spectrum's blue component doesn't affect the sound.)

Pictures are rendered into sound using a variety of sound sources, including a variable-waveform oscillator, a sample player, a multisampled instrument, a pseudo-FM synth, and a granular synth. The Image Synth can be approached as a kind of glorified piano roll - style note editor or as a sound canvas for painting and analyzing sound files. The view you adopt during any session will greatly influence how you use MetaSynth's many graphic-editing tools, and it will also affect your choice of rendering sources. I'll describe both approaches in this article.

Analyze this. Analyzing sound files is a good way to get acquainted with the Image Synth. The process is as simple as loading a sound file into MetaSynth's Sample Editor, switching to the Image Synth, and selecting "Analyze current sound . . . (n)" from the File submenu (shown as the disk icon at the top left of the Image Synth; see Fig. 1). MetaSynth performs an FFT (Fast Fourier Transform) analysis of the sound file and displays an editable sonogram. If you then render this image using sine waves, you will get a resynthesized approximation of the original sound file. The quality of the resynthesis depends on the resolution you have chosen for the horizontal (time) and vertical (frequency) dimensions of the Image Synth.

Fig. 1 shows the Image Synth Palette (top) together with an analysis of the spoken word "welcome" at three resolutions. The Image Synth's frequency resolution is controlled by the scale selected from the Map menu at the top right of the window. From left to right, the three analyses were made using the Semitone, Micro8 (eight divisions per whole tone), and Micro16 scales. As seen in the pictures, higher resolution (that is, more frequency bands) results in a finer analysis of the sound file.

Getting around in the Image Synth. The Image Synth's tools and pop-up submenus are arranged along the image area's borders. Some of the tools are invoked by a single mouse click, and others by clicking and dragging in one or two dimensions. In the latter case, the Shift key constrains dragging to the horizontal, and the Shift+Option key combination constrains dragging to the vertical dimensions. Double-clicking on many of the drag tools provides access to additional parameters, and many Image Synth functions can be invoked by keyboard commands.

The Image Synth's time resolution is controlled by the Tempo and Duration dialog, which is opened by double-clicking on the watch icon on the top border. Because the sound file's length is measured in samples and the Image Synth's width is measured in pixels, the natural unit of time resolution is samples per pixel (SPP). Fewer samples per pixel means higher time resolution but requires correspondingly wider pictures for the same size sound file. (Once you select the picture width using the Size x menu, MetaSynth can calculate the SPP setting required to fit the sound file into the Image Synth.) From left to right, the three analyses in Fig. 1 were made using 343, 171, and 85 SPP settings.

To better illustrate the differences in resolution, the pictures at the bottom of the figure show only the first few octaves of the analysis. There is, of course, a good deal of information above this, as you can see from the full view of the Image Synth (top of Fig. 1). In an analysis using microtonal scales, typically you will find large vertical clusters of pixels because the analysis smears over adjacent frequencies. During rendering, if the playback phase relationships are exactly right (or better said, exactly wrong), clipping can result. The solution is to lower the amplitude of the rendering input source. If this source is MetaSynth's Wave Table Palette, simply use the Scale Tool (third button from the left) while holding down the Shift and Option keys.

Sound file resynthesis. Although resynthesizing an analyzed sound file may seem pointless, many interesting variations can be produced by applying the Image Synth's graphic tools to the image before resynthesizing. A simple starting point is to change the Image Synth's resolution by altering the tuning, Frequency Map, or SPP setting. For example, try analyzing a short speech clip with the Micro8 scale. Then change the scale setting to Micro12 and render the clip at twice the SPP. Also try to set the tuning (using the tuning-fork icon on the top-right border) an octave higher.

Another effective approach is to change the input source used for rendering the picture. Again, analyze a speech clip. With the sound file still in the Sample Editor, render the image at four times the SPP using the Sample Granular input source. (The Sample Granular method is new in MetaSynth 2.7.) You'll hear four somewhat different versions of the original clip because the grain must cycle through the sample four times to render the elongated picture. As an alternative, choose Narrow Noise from the Sounds menu before rendering the image with the Sample Granular input source. This produces a kind of outer-space radio effect, but the words remain intelligible. Finally, try applying some of the filters from MetaSynth's Filter Palette to the noise sample before rendering.

Pitch and time effects. MetaSynth has three effects groups for quick picture modification (that is, modifications that you don't have to paint in with the brush tools). The Hot Filters that are invoked using the buttons along the Image Synth's lower-right border are time based; they modify horizontal aspects of the picture. The effects available on the Pitch and Harmonics submenu (on the left border with the notes icon) are pitch based and modify vertical aspects of the picture. Finally, the effects on the Process submenu (on the left border with the water-drop icon) modify the picture in both dimensions.

Because you're working graphically, you're not restricted to using any of these effects in their intended dimension. For example, applying a pitch effect to time is as simple as rotating the picture 90 degrees. Of course, there's a tool for this, too. The button on the bottom border with the quarter-circle icon can be used to rotate pictures by dragging or by using numerical values. One effective application of this technique is to use the Add Harmonics pitch effect to generate an accelerating echo or "bouncing ball" effect.

This process works best with short, percussive hits with a little silence at the end. (If the sound file you're analyzing has no silence at the end, increase the SPP setting by about 20 percent after fitting the Image Synth to the sound file.) The first step is to multiply the picture's horizontal dimension (Size x) by four and set the vertical dimension to match. Next rotate the picture 90 degrees counter-clockwise by double-clicking on the rotate tool and entering "-90" for the rotation amount with "Wrap around" turned off.

Echoes are created by adding harmonics for each pixel in the picture. Harmonics are integer multiples (2:1, 3:1, and so forth) of the fundamental frequency. New pixels need to be inserted on the rows that represent those frequencies in the same columns as the original pixels. The Image Synth's Add Harmonics process does this, and because it works relative to the chosen Frequency Map, you can spread out the harmonics by choosing one of the micro scales (Micro32 and Micro50 are good choices). Once you've chosen a scale, select Add Harmonics from the Pitch and Harmonics submenu. If you hold the Option key while doing this, you will get more harmonics, hence more echoes.

The last steps are to rotate the picture 90 degrees clockwise and restore the original Frequency Map. The harmonics now become echoes of the original pixels, and they die out because the Add Harmonics process adds harmonics with decreasing intensity. An interesting variation is to rotate the picture a little more or less than 90 degrees clockwise at the end - this will make the echoes rise or fall slightly in pitch.

Stealing the groove. A nice thing about analyzing a rhythmic sound file is that its groove comes along for free. Fig. 2 shows the steps taken to analyze the groove of a 2-bar bass loop. At the top, the bass loop is analyzed in semitones at 32 pixels per beat. Low resolution is fine and even preferable here, because the goal is not to resynthesize the sound, but rather to find the accents.

This sound file is particularly easy to analyze, because though most of the action is in the bottom register, the string slap leaves a trail in the higher harmonics. Therefore, you can throw away the bottom part of the image and still retain the groove. This can be seen in the second frame from the top in the figure. In some cases, you may need to manually clean up the picture a little. The "Shorten" Hot Filter and "Noise Filter" options on the Processes submenu are helpful for that purpose.

The next step is to use the "Max" Hot Filter to turn the beats into bars, as shown in Fig. 2's third frame. When you apply this particular effect, the program determines the brightest pixel in each column and gives all of the pixels in that column the same brightness value as the brightest pixel in their respective columns. (The Pitch and Harmonics submenu has a similar function for pitch.) Notice that the groove bars are rather narrow. You can use the "Reverb" Hot Filter with a small grid size to expand the bars and make a more effective filter (Fig. 2, bottom).

Once you have a groove filter, you can use it for filtering pictures or sound files. In either case, because the filter is imposing the groove, the effect works best on nonrhythmic source material. Try copying the groove filter to the Image Synth's clipboard (type the C key), then opening MetaSynth's Filter Palette and pasting the filter (type V). Change the original loop to noise in the Sample Editor, and apply the filter to the resulting noise sound file. You might mix this with the original sound file or use it as a source for MetaSynth's Convolve and Formant Filter Morph operations. As an alternative, you could apply the groove bars to one of the other filters in the Filter Palette. Just type Shift+8 to filter whatever is in the Filter Palette with the groove-bars image on the clipboard.

Vocoder-like effects and the Filter Palette. A variety of DSP techniques, in effect, impose the time-variant frequency spectrum of one sound file (typically speech) on another sound file (usually pitched or noise). MetaSynth provides four such processes on its Morph menu: Cross Convolve, Formants Filter, Osc Bank PhaseVocoder, and Convolve. You can also produce a wide variety of these kinds of effects by using filters in either the Image Synth or the Filter Palette. The use of groove filters, described previously, is one way to produce such an effect.

MetaSynth's Filter Palette is similar to the Image Synth, except that it has fewer tools (for example, no Hot Filters), and the window size, Frequency Map, and tuning are fixed. The window size is 256 (horizontal) by 128 (vertical) pixels, and the Frequency Map is in semitones tuned to the pitch A1. Keeping these limits in mind, you can transfer pictures freely between the Image Synth and the Filter Palette. Just make sure that the Image Synth's parameters are set to match the Filter Palette's. The active filter library is shared between both windows so that when you open or change a filter file in one, the change is reflected in the other.

Using the Filter Palette for vocoding is a simple matter of transferring a sonogram of the source material to the Filter Palette and applying it to any sound file with sufficient harmonic content to produce intelligible results. As a quick example, set the Image Synth's dimensions to match the Filter Palette, load a short speech sound file into the Sample Editor, fit the Image Synth duration to match the sound file, and analyze the sound by typing N. (The reason to analyze the sound file in the Image Synth is that analysis in the Filter Palette produces a reference template rather than a filter.) Now copy the sonogram to the Image Synth's clipboard and paste it into the Filter Palette. With the sound file still in the Sample Editor, select "WhiteNoise" from the Sounds menu. This creates a noise sound file of the same length. Apply the filter by clicking on the Filter Palette's Synthesize button (the Mac icon on the bottom border).

As with resynthesis, the ability to graphically manipulate the pictures used as filters distinguishes this process from run-of-the-mill vocoding. The five tools at the bottom left of the Filter Palette - scaling, rotation, offset, contrast, and displacement - are particularly effective. For example, the scaling tool used horizontally provides time compression and stretching without formant or pitch shifting; the offset tool provides pitch and formant shifting without affecting time; and the displacement tool provides a variety of effects from simple vibrato to a total mangling of the sound. Each of these can be used by clicking and dragging in one or two dimensions or by double-clicking to enter parameters numerically.

Stereo and color. The images shown here have all been black and white, which means the analyzed and rendered sound files have all been mono. MetaSynth can also analyze and render stereo files when the Image Synth is in stereo mode. (To switch between stereo and mono modes, click on the sample icon in the middle of the top border.) If you analyze a stereo sound file in mono mode, MetaSynth analyzes the left (red) channel.

Analyzing mono sound files is faster, and for things like groove analysis, it is generally a better choice. Once you have analyzed a file in mono, however, you may want to convert it to stereo in the Image Synth. This is handy for rendering stereo effects and creating templates in the silent blue channel.

When the Image Synth is in stereo mode, a blue-grid submenu appears on the top border. You can use this to create various horizontal and vertical grids quickly, and you can paste the clipboard's contents here. In the case of a groove analysis, setting the grid size to 16 and typing G produces blue gridlines at eighth-note intervals. The gridlines give you a guide against which you can compare the groove to straight eighths. If you copy the groove analysis to the blue channel, you can utilize it as a painting template. You can also use the blue-grid submenu to apply the image on the blue channel as a filter.

A little of this, a little less of that. The Image Synth is easily configured for additive synthesis, and with the use of both graphic and audio filters, it produces a unique collection of synthlike samples and multisamples. Fig. 3 shows Image Synth analyses of six typical analog synthesizer sounds: unfiltered sawtooth and square waves (top row), sawtooth with decay and attack filter envelopes (middle row), and sawtooth with LFO applied to volume and filter cutoff (bottom row).

These analyses were made with a custom Frequency Map in which each row's frequency is a harmonic of the bottom row's frequency. Rendering these with sine waves gives a reasonably accurate additive replica of the original sound. However, because the phase relationships between the harmonics can vary with each rendering, the resulting sound does not have the same graphic waveform. A sawtooth rendered additively, for example, does not look like a sawtooth in the Sample Editor.

Add 'em up. For pure additive synthesis, an exponential Frequency Map should be used. Other custom scales can be used for additive synthesis, but typically they should contain at least as many steps per octave as the picture height, and the steps should bear some harmonic relationship to each other (more on this in a moment).

Building an additive sound in the Image Synth gives you control over each individual sine component's amplitude and pan envelopes. For each component, start with an empty window and use the Line brush to create a line at the desired harmonic (that is, row). Then use filters or the Filter brush to color and contour the intensity of the line. Finally, save the line as a preset, clear the Image Synth, and work on the next harmonic. Saving the individual components as presets lets you mix and match them later to create any number of sounds. You can do this by recalling the first preset, copying it to the clipboard (type the C key), recalling the next preset, adding the clipboard data to it (type E), copying the composite result to the clipboard, recalling the next preset, and so on.

A faster approach is to work on several rows at a time. You could do this by creating a horizontal blue grid and using it as a filter to select groups of rows. To quickly create the grid, activate the blue channel, set the Hot Filter Grid Size to the desired row spacing, type G to create a blue grid with this spacing, and use the Rotate tool to rotate the grid to horizontal. If there is unevenness in the grid, drag the Contrast and Luminance tool to the right to eliminate the contrast. Next select the red and green channels, type I to invert the empty Image Synth to full yellow, and select "Filter with Blue" channel. As before, use filters and the Filter brush to contour the remaining lines, then save the image as a preset. After you've filtered the full-yellow image with the blue channel, use the Arrow keys to move the remaining lines up and down as desired to create the other harmonics.

Fig. 4 shows an example with blue gridlines spaced eight rows apart, harmonics added on two rows within each section, and the fundamental in full yellow at the bottom. An increasing tremolo was added to the right half of the image. You can render this image, then use the resulting sample in the Sample Editor as a sound source. (See the sidebar "MIDI Sequencing with MetaSynth and Xx" to see how you can use standard note sequences to trigger sounds.) However, the standard disadvantages apply: the sample's length and tremolo speed will change when transposed, and the upper harmonics will become unnaturally shrill at higher pitches. Building a multisample will produce better results for use in either a MetaSynth Instrument or another sampler.

To build a multisample, simply render the image using different tunings. Spacing by major or minor thirds usually keeps the speed changes of any motion (for example, tremolo) within acceptable limits. Additional lowpass image filtering for each octave change in pitch knocks down the shrillness. Finally, appending the tuning pitch to the end of each sound file's name will facilitate building instruments from the multisamples in MetaSynth and many other samplers. In MetaSynth, selecting "Build Instrument" from the Instruments menu - then selecting any of the sound files - results in all of the samples being loaded, set to the correct root pitch, and correctly mapped relative to the semitone Frequency Map.

Beyond the norm. You can produce interesting and unusual samples by using custom scales and some of MetaSynth's more graphic tools. One of the most interesting ways to create a custom scale that works well for additive synthesis is to analyze the harmonic spectrum of some sampled sound. This technique works with any sound, but pitched sounds work best. Once you have loaded a sound file, select a region in the sustained part of the sound and choose Instant Spectrum from the Morph menu. MetaSynth extracts the harmonic spectrum of a small clip at the selection's beginning. To import the spectrum file as a custom scale, first select Custom Scale from the Frequency Map menu. In the Custom Scale window, hold the Option key and click the Open icon (the disk with the upward-pointing triangle). Use these custom scales to build additive samples as you would with the Exponential scale.

The Displacement tool stands out as one of the more creative tools for image manipulation. This tool moves pixels in the Image Synth based on the intensity of pixels in the displacement-map picture. Fig. 5 (middle frame) shows the result of applying the displacement map shown in the top frame to the picture shown below it. Displacement can be applied with or without automatic smoothing; smoothing is usually desirable in this context.

The Filter Palette is another place to turn for sculpting unusual sounds. One thing to remember is that graphic filters can be applied to any selected part of a picture. The bottom two frames in the figure illustrate this - first, semicircular filters were applied to the right and left quarters, then a phasing filter was applied to the bottom horizontal section, and finally, a lowpass filter was applied to the middle horizontal section. The resulting picture produces various speechlike effects when rendered with harmonic-spectrum Frequency Maps. You can also use the Hot Filters or groove filters, like the bass groove described previously, to add rhythm and motion to the sound.

When applying filters, keep in mind that any MetaSynth filter can be applied in two ways: graphically to a picture before rendering or as a 128-band EQ to the sound file in the Sample Editor. Applying a filter graphically is generally harsher because its effect is absolute - filtered-out pixels make no contribution to the sound. Audio filtering, on the other hand, produces resonant peaks and valleys.

EVEN MORE TOOLSMetaSynth provides a broad range of DSP tools. Some of these are found on the Transform and Morph menus, and among these some have their own unique graphic interfaces. A typical example is the CrossConvolve window shown in Fig. 6.

Cross convolution is one of the vocoder-like effects mentioned earlier. It operates on two sound files by performing an FFT analysis of each and multiplying the resulting spectra. The effect is similar to analyzing both sound files in Image Synth, then using one image to filter the other. However, cross convolution uses much higher resolution. Because image filtering is simply pixel-by-pixel multiplication, it is symmetric. It makes no difference which picture filters the other. The same is true of cross convolution.

MetaSynth's approach to cross convolution provides independent envelopes for mixing each of the source sound files with the cross convolution. Some familiar graphic tools are provided across the bottom of the envelope window, and you will find wave-shaping tools along the right edge (see Fig. 6). In the figure, the sound file in the Sample Editor is faded in (green envelope), while the sound file on disk is faded out (red envelope). The cross convolution fades in and out with a bell-shaped curve (blue envelope). To hear the result of cross convolution by itself, flatten the red and green envelopes at the bottom and flatten the blue envelope at the top.

Loop Bassix. Here's an example of how you might apply cross convolution along with the MetaSynth processes described above to produce variations of a bass loop. First load a bass loop into the Sample Editor, fit the Image Synth duration to it, and analyze the loop to make a groove filter (see the section "Stealing the groove"). Save the groove filter in the Filter library.

With the bass loop still in the Sample Editor, select a small clip with as much harmonic content as possible and make an Instant Spectrum from the clip. Then select Custom Scale as the Frequency Map and import the spectrum as a scale. (To have the spectrum recognized as a scale file, hold down the Option key when clicking on the import-scale button - the disk icon with an upward-pointing triangle.)

Next clear the Image Synth and paint or recall a picture suitable for additive synthesis (see the section "Add 'em up"). Ensure that the Image Synth's duration still matches the bass loop's length and that its Frequency Map is set to the spectrum scale from the previous step. Then either render the picture and filter the resulting sound file in the Filter Palette or graphically filter the picture before rendering. Save this sound file to disk, so you can recall it for each application of cross convolution.

The final step is to cross-convolve the sound file with the original bass loop a number of times using different envelopes. To keep the various convolutions consistent and have the original bass loop high in the mix, set the red envelope to a horizontal line fairly high in the window and modify only the green and blue envelopes for each variation.

Once you have a number of variations on the original bass loop, you can use them in your audio sequencer or sampler, or you can use MetaTrack to create composite loops (see the sidebar "Audio Sequencing with MetaSynth and MetaTrack"). You can also quite easily build composite loops directly in the Image Synth.

Taking note. The simplest way to gain access to all the loop variations as rendering-input sources is to first build an Instrument from them. (See the end of the section "Add 'em up" for an Instrument-building shortcut.) Because you probably don't want to play the loops at different pitches and speeds, they should be mapped to consecutive notes. This will restrict you to working in a vertical region of the Image Synth, which has a height determined by the number of loop variations. Use the blue channel's Line tool to mark this region's upper and lower boundaries.

Fig. 7 shows a note sequence for combining six bass-loop variations from the previous section. To create this sequence, the picture width was set to 128 and the duration was fit to the original bass loop, then multiplied by 8. As a result, a line 16 pixels wide will play the loop once. The Note brush was set to a width of 16/1 so that one click enters a 16-pixel line one row high. The Snap Grid (the green grid on the left) was turned on and set to 16 so that entered notes will be quantized to a 16-pixel grid. The horizontal blue gridlines indicate the outside boundaries of the Instrument's note map.

The long notes were painted in first with a yellow Note brush. They were then colored by applying a colorizing filter and faded in and out using a black Filter brush. Rendering the file using the Instrument as the input source at this point would cause each loop variation to fade in, pan across the stereo field, and fade out. The short notes at the top (four pixels wide) were entered later to double the beginning of each loop with a different variation. Each one plays the loop's first quarter at the opposite side of the stereo field from the longer line at the identical horizontal position. Of course, you can do a great deal more with the Image Synth's brushes. This example indicates the complete control you have over volume and pan envelopes and shows how quickly you can enter note sequences.

JUST FOR EFFECTIn addition to the effects on the DSP and Transform menus, there are 14 effects accessible from MetaSynth's Effects Palette (see Fig. 8). A pair of these - Harmonics and Inertia - use the waveform's harmonic spectrum in the Wave Table Palette to color the sound file. Four others - Stretch, Grain, Shuffler, and Harmonize - use granular-synthesis techniques. Here's an example using a couple of the more unique effects to process the composite bass loop from the previous section.

The Inertia effect simulates sympathetic vibrations based on the waveform's harmonics in the Wave Table Palette. Used sparingly, this can add subtle and effective resonances to any sound. Applying it to a sound file's different segments using different waveforms in the Wave Table Palette adds variation to an otherwise repetitive loop. In the context of the bass-loop composite, try applying different Inertia to each quarter (for example, four bars) of the sound file. Save the sound file for reuse later.

The Grain effect breaks the selected portion of the sound file into grains that are then fed into and played out of the "grain engine" at different rates. In the process, the playback order can be randomized to varying degrees. This can produce everything from subtle time-shifting effects - when the in and out rates are close and little or no randomization is present - to complete mangling of the sound file with widely varying in and out rates and heavy randomization.

Try processing the Inertia-processed composite bass loop in two halves as follows: set the grain size to quarter-note length (if you have eight two-bar loops as above, this amounts to dividing the entire sound file length by 64). Set the Input Step and Output Step to one-quarter of this value and set Randomization to zero. Audition the loop, and you'll hear little if any difference from the original. Now select the first half of the sound file, increase the Input Step setting slightly, and apply the Grain effect. Next select the sound file's second half, decrease the Input Step to as much below the Output Step value as it was above it, and apply the Grain effect. Finally select "Mix 50%" from the Morph menu and mix the sound file with the Inertia-processed sound file on disk.

Using Cross Convolution, Inertia, and Grain one after the other as described here may be a bit over the top, but it shows that you can get intelligible results even with that much processing. Nevertheless, when it comes to effects processing - in MetaSynth as with any application - less is usually more.

TWO STEPS FORWARDThis article began with sound-file analysis, which produced sonograms in the Image Synth. This is an excellent way to get a sense of what kind of graphic is needed to produce a specific type of sound. Sound-file analysis is also a powerful tool for rhythm, pitch, and harmonic analysis. In the section "Analyze this," you glimpsed several analysis and resynthesis processes that are greatly facilitated with graphic tools, and in some cases impossible without them. Next I looked at sound painting using the model of additive synthesis. In addition to using the standard overtone series, I employed custom-designed harmonic structures, including those derived by analyzing other sounds. I then used both graphic- and audio-filtering techniques to modify the sounds before assembling them as multisamples for use in MetaSynth Instruments or other samplers.

Finally I explored a few of MetaSynth's vast array of DSP processes, which range from the standard to the exotic. Here I used a subtle mix of additive synthesis and cross convolution to produce variations on a bass loop. I then strung the variations together and processed the result with resonating and granular effects. A functional bass loop with the same pitch and rhythm remained, but with a lot of harmonic variation.

MetaSynth is a deep and complex program, and this article is a bit like skipping a stone across a pond's surface. Although I've touched on the program's basic elements, much time can be spent exploring any feature without mastering it completely. As seen here, you can also get usable (and unique) results quickly. So don't be put off by the program's complexity; roll up your sleeves and get started.

MetaTrack is a 16-track audio sequencer companion to MetaSynth that is included on the MetaSynth Studio CD. Like Xx (see the sidebar "MIDI Sequencing with MetaSynth and Xx"), it will not replace your digital audio sequencer, but it is specifically designed to make creating montages of rendered presets a seamless process. Each track has independent pan and volume control as well as one optional insert effect. Your CPU's speed and your hard disk, as well as the number of insert effects used, determine how many tracks can be auditioned. Because the final mix is not done in real time, you can always mix a full 16 tracks including effects.

MetaTrack and MetaSynth share preset libraries, and MetaTrack searches for split-stereo sound files with the default MetaSynth names "Sound##.L" and "Sound##.R." MetaTrack also lets you assign any sound file to any library preset. However, when it does this, it creates a sonogram of the sound file and replaces the preset graphic with this sonogram. Because these changes are automatically saved to disk, the original graphic is lost in the process.

For this reason, it is advisable either to create a new preset library or copy an existing one for use in any MetaTrack project. It is also a good idea to create a preset and associated sound file for each sound used in a montage. For example, if you render a preset and apply a filter or effect to it, then wish to use both versions, you should create a new preset for the processed version. Finally, keep all files related to a montage - preset library, sound files, montage files - in a separate project folder.

Fig. A shows the MetaTrack window. The tab below the menu bar contains the preset library. Below it, six audio tracks are displayed, and below those are shown various transport, display, and editing controls. When you click on the speaker icon to the right of the library tab, the associated sound file plays. If no sound file with the default name is found, you will be prompted to select one. You can also double-click on the preset to reassign its sound file.

You fill in the audio tracks in MetaTrack by dragging presets onto them. The presets are snapped to an invisible grid measured in pixels, which you can set in MetaTrack's Preferences. By default this is set to 32 pixels (that is, one standard MetaSynth beat). You can also set the size of a measure in pixels (128 by default) and the tempo in samples per pixel (SPP). When you open a preset library, MetaTrack's tempo is already automatically set to that of the first preset.

If you have MetaTrack and MetaSynth running at the same time with the same preset library (the standard operating procedure), several shortcuts can facilitate creating and editing a montage. From MetaSynth, the F9 function key will cause the picture in the Image Synth to be rendered, saved to disk, and saved as a preset replacing the currently selected preset and sound file. In the other direction, MetaTrack's MetaSynth menu contains several options for processing the selected preset in MetaSynth, including opening it for editing, rerendering it, and applying MetaSynth filters or effects.

Xx is a 16-track MIDI sequencer included on the MetaSynth Studio CD. Although it is fairly rudimentary in some respects and won't replace your full-featured MIDI sequencer, it has some interesting and unique features for automatic pattern generation and manipulation. Its simplicity, combined with easy linking to both Beatnik and QuickTime Musical Instruments for playback, also makes it an excellent MIDI sketching tool.

The real reason for Xx's existence, however, is to facilitate the transfer of MIDI files to and from MetaSynth. These can be MIDI files created directly in Xx or imported as Standard MIDI Files (SMFs) from some other source. The more obvious use - transferring MIDI files to MetaSynth - is a great improvement on entering music note-by-note with MetaSynth's Note Brush tool. The reason to do this at all is, of course, to use specifically MetaSynth techniques to turn the MIDI file into a sound file. These might include anything from graphically editing individual notes to applying large-scale horizontal (time) and vertical (pitch) modifications, such as adding echo, reverb, or harmonics.

Converting graphics to MIDI files (that is, transferring pictures from MetaSynth to Xx) can also be useful, though the considerations about rendering high-density graphics also apply to converting them to MIDI files. In short, you can produce a wide variety of note patterns by using some of MetaSynth's brush tools (albeit sparingly), then converting the picture to a MIDI file.

Xx was designed to convert graphics to MIDI files, which makes using it with MetaSynth extremely simple. Fig. B shows Xx's Export to MetaSynth dialog, where you'll notice a number of settings that affect the transfer. The Export menu at the top left lets you decide how the sequence will be exported: as a picture file, as a new preset l library with or without separating tracks, or as presets added to any existing preset library you select. Xx uses 16 samples per pixel and assigns Middle C (262 Hz) as MIDI note number 48 by default. Because the Image Synth's default resolution is 32 samples per pixel and MetaSynth Instruments assign Middle C to MIDI note number 36, Xx automatically compensates when exporting presets by halving the tempo in the Image Synth and transposing down an octave. When you export as a picture, however, only the graphics are exported. You will need to make any time and tuning corrections manually.

The Harmonic maps at the lower left of Fig. B let you differentiate between how high- and low-velocity notes are translated. For example, you can add harmonics to higher-velocity notes. This works like a velocity crossfade rather than a velocity switch. The Poly/Mono switch determines whether only the active Xx track or all tracks are transferred. Remember that the Export menu determines whether multiple tracks are exported as separate pictures. The Controller settings on the right affect the velocity sensitivity and the transfer's attack/release characteristics.

There are fewer choices when importing pictures in Xx (note that this is not restricted to pictures generated by MetaSynth). The options are to separate the red, green, and blue channels to individual Xx tracks or set an intensity threshold below which pixels will not be imported. The threshold applies to lines rather than individual pixels; if any pixel in a line is above the threshold, the whole line is imported.