Transforming Sound With Ceres3

Ceres3 is a spectral domain editor for audio files. Its display represents the frequency content of a sound as it changes over time, and the program's toolkit provides the means for directly editing and transforming that content. Ceres3 is a fantastic tool for studying, editing, and creating sounds, and in this tutorial we'll take a look at what can be done with some of Ceres3's powerful Transforms.


The examples here were made under certain conditions. I used a 29-second monaural AIFF file (Ceres3 reads and writes only mono AIFF soundfiles), and I started the program with this command sequence:

	ceres3 4096 4096 128 800

The command options set the FFT and window sizes, the window step factor, and the display width. Higher values for the first three options will result in higher resolution analysis and a much larger analysis file.


Spread applies a granulation effect to the sound. Figure 1 displays the spectral content of the unaltered file, a reading by Ezra Pound :

Figure 1: Ceres3 with soundfile loaded
(Click here for large view)

You listen to this file here.

After selecting Spread from the Transforms menu we apply these values to the transform parameters:

Figure 2: Spread parameter settings

Figure 3 shows the results:

Figure 3: The soundfile after applying the Spread transform
(Click here for large view)

The actual output sound is an audio analog of the visual display. At first only a few scattered grains are heard, but by the mid-point of the time-scale the sound is rapidly coalescing into something more like the original file. By the end-point the sound is (more or less) identical to the original.

The transform's parameters are rather sensitive, and you may find that only a rather narrow range of values is useful for some parameters. For example, altering the random spread yields relatively subtle differences within a range from .50 to 2.0, but applying a similar range to the multiplier has a dramatic effect. The multiplier "advances" the granulation into the soundfile, i.e., at .25 the granulation occurs over only the first quarter of the file, at .50 it reaches the halfway point (as seen in Figure 3), but at .75 almost the entire file is granulated.

I leave it as an exercise for the reader to discover the effect of varying the amplitude sensitivity. Note also that this Transform can be dynamically controlled by Ceres3's function curves (see the Edit/Extract dialog) and that the effect can be restricted to within a delimited area (see Settings/Edit Display).


Figure 4 shows some useful settings for the Exponentiate transform:

Figure 4: Exponentiate transform settings

These values will have a dramatic bending effect on the sound, as can be seen in Figure 5 and heard here.

Figure 5: Soundfile after Exponentiate transform
(Click here for large view)

This transform is especially sensitive to values for its multiplier and multiplying factor; however, the frequency exponent can be adjusted over a fairly wide range and a control function can be applied for dynamic modulation of the exponent value. Like Spread, the transform's effect can be delimited to a restricted area of the sound. Finally, the initial frequencies produced by the example may be filtered out or shaped by the Ceres3 paint brush. Yes, you can directly edit the spectral content with a resizable paint brush tool, and yes, it is very cool.

Combining Transforms

Our last example will combine three transforms to create a rather interesting "musicalizing" effect from the spectral content of our original soundfile.

First apply the Sieve and the Blur tramsforms with their default values. These actions will result in the display shown in Figure 6:

Figure 6: Soundfile after Sieve and Blur transforms
(Click here for large view)

Next open the Settings/Pitch Grid dialog and set its values to those in Figure 7:

Figure 7: Pitch Grid settings

From this point we will apply the Move To Pitch Grid transform (using the defaults) over three sections of the soundfile, changing the Pitch Grid settings each time to create a sense of harmonic motion in the resynthesized sound. The first part was set for Messiaen's third mode of limited transposition (base frequency at 60 Hz), the second section was set for the odd-numbered members of the harmonic series starting from a base frequency of 90 Hz, and the last part used a pitch grid for a heptatonic altered scale with a base frequency of 120 Hz. Figure 8 shows the results, you can hear them here.

Figure 8: Soundfile after combined processing
(Click here for large view)

Going Out...

You can keep piling on transform after transform (the Mirror and Spectrum Shift effects might be interesting), and Ceres3 contains many other interesting aspects, but I must stop somewhere. Fortunately the program's interface is well-designed and invites experimentation, so you will have no trouble making some joyful noises of your own.