MORPH: Detailed Information

by Archer Endrich

Key Features of the CDP MORPH set

• The morphing is achieved by the interpolation of spectral data (frequency content) over time.
• CDP's MORPH deals with the whole spectrum at once. Other morphing software, such as SNDAN, takes more account of individual partials and might be more effective with pitched sounds.
• Frequency and amplitude start points can be set independently.
• MORPH MORPH contains a stagger parameter by which the start of the 2^nd file can be offset. This makes it possible to customise the location of the morph region.

Some possible musical results with the CDP MORPH programs

• Trevor Wishart initiated CDP's morphing software – and the whole of the spectral domain program set – when he sought to duplicate the famous human voices into bees morph which began his composition, Red Bird. This passage of just a few seconds took two weeks to complete in the 'classical' tape deck studio. Red Bird is filled with all kinds of sonic transitions and a classic work in the genre. His initial spectral programming was done at IRCAM, with the assistance of programmers there.
• The morphing process is easiest to implement when the two input sounds are similar and at about the same pitch level.
• Time-varying pitch transposition can help to provide some middle ground between sounds at distinctly different pitch levels, which is often the case with voices. But due regard must be made for the ease with which the ear perceives vocal changes to be unnatural or weird.
• Sufficient time must be allowed for the ear to observe the transition. Sometimes time-stretching can help to provide the time needed. A time-varying stretch factor makes it possible to keep the first part of the sound as is (stretch=1) while stretching out the remainder.
• The MORPH BRIDGE and MORPH GLIDE functions can make very dramatic continuous transitions when the spectral content of the two windows selected is strongly dissimilar.

Other forms of morphing in CDP (transitions and reshaping one sound with the characteristics of another)

• The GLIDE function enables gradual transition between the spectral contents of any two single analysis windows in the input analysis file.
• The BRIDGE function does the same, but using single analysis windows drawn from two different analysis files.
• CDP's broad range of sound transformation functions make it possible to prepare sounds for morphing. This makes assists in achieving difficult morphs (overcome differences in the sounds) and to fine-tune the psycho-acoustic perception of the morph itself.
• FORMANTS VOCODE transfers the spectral envelope of the second file onto the first (='cross-synthesis'). This works best when the spectral envelope has a lively profile. This profile, based on the amplitude level of the frequency content, has the effect of reshaping the first sound with the timbral characteristics of the second.
• COMBINE CROSS makes a gradual transition from the amplitude of the first spectral envelope to that of the second. It does this by making a weighted not a full substitution, so it is not as thorough a process as a morph. However, it can be a useful first step in moving between significantly different timbral characteristics, thus preparing the two sounds for a full morph.
• Combinations which involve transposing the extracted pitch trace of one sound with that of another provide a different approach to sonic reshaping. For example, if a wide pitch trace is transposed by the ratios of a narrow one, the first sound then seems as if it is being squeezed through a narrow hole. See REPITCH COMBINE.

General observations

• Morphing with sounds generally takes longer than morphing with visual images. The ear is acutely sensitive to the minutest change and is not easily fooled.
• When there are significant amplitude envelope differences between the sounds (e.g., the second one is rhythmic), one technique is to impose the amplitude envelope of the second sound on at least a part of the first before morphing (ENVELOPE REPLACE). Blurring might also be used to minimise various peaks in one or both sounds.
• A smooth morph may not sound any different from a cross-fade. There are times when one wants to feel the first sound being pulled out of shape and reformed as something new. Determining the nature and degree of the mid-point distortion takes considerable thought and, more often than not, experimentation.
• A judicious use of the stagger parameter can help to place the morph section after a portion of unaltered first input and before the end of the second. This helps to give the ear time to hear the before and after, and also results in a longer output soundfile.
• Some sense of disturbance or warping of the first sound as the morph gets underway can add to the drama and effectiveness of the morph: you can hear that the first sound has gone awry and don't yet know where it will end up.
• I have been experimenting with amplitude and formant envelope transfers, blurring, transposition, spectral crossing and some aspects of waveshape distortion to achieve this sense of disturbance, with variable degrees of success.
• It is best not to underestimate the challenge of achieving effective audio morphs.

Last updated: 4 December 2003