Return
to Main Index for the CDP System.
Return
to Groucho Index
infile – input sound to process
outfile – resultant soundfile
join_time – time in infile where join-cut is to be made
splice-length – length of the splice, in milliseconds
I call the outputs of this process (learnt from Denis Smalley), fugu sounds. There is a particular Japanese fish delicacy, the fugu fish, which has a poisonous liver, but tastes best the nearer to the liver you slice the fish (!). Fugu sounds are made using an attack-resonance source (a sound with a sharp attack which then fades away to nothing). A Reverse copy of the sound is made and then spliced onto the original so the sound now grows from nothing until it reaches a peak of loudness and spectral brightness and then fades once again to zero.
If the crossfade is made just before the peak is reached, the sound is less loud and bright in the centre, and in fact a whole sequence of such musically related sounds can be made, each of different loudness/brightness in the centre. This process allows such sounds to be made in a single pass.
End of EXTEND BAKTOBAK
Return to List of functions for interacting
spectra at top of this file
Return
to Main Index for the CDP System.
Return
to Groucho Index
infile – input soundfile
outfile – output soundfile, with repetitions
segdur – duration of segments
repets – number of times each segment is repeated
[-s] – option to have outfile try to stay synchronised with the infile
segdur may vary over time
EXTEND DOUBLETS is a 'slice' function, like the ones we are familiar with in the visual realm. The difference, here, in the temporal realm, is the repetition parameter. We specify the length of the segments (slices) and the number of times it repeats.
What we hear depends, as usual, on the sonic material. With voices or conventional music, the effect will be like the needle getting stuck on a vinyl record: a short passage repeats. With more complex sonic material, we would get a pulsing, mechanical effect. Especially note that the length of the segment (segdur) parameter can vary over time.
Here are a few ideas to stimulate the imagination:
- a speaker repeats a phrase for emphasis
- a speaker repeats his phrase, adding more each time
- a sound repeats, unfolding as it does so
- extended time-varying extensions of a sound or syllable
- mechanical throbbing
- extending the material of a sound prior to using other sound transformation processes
End of EXTEND DOUBLETS
Return to List of functions for interacting
spectra at top of this file
Return
to Main Index for the CDP System.
Return
to Groucho Index
Splice segments of source file end-to-end; the start times of the segments in the source file are chosen by a 'drunken-walk' through the source file. In Mode 2 the source file plays soberly at holds.
1 Drunken walk
2 Play soberly at holds, with lower and upper limits of sobriety
infile – input soundfile to process
outfile – output soundfile
outdur – total minimum duration of output soundfile (seconds)
locus – time in infile at which the drunken walk occurs (seconds) – this location can move through the source
In breakpoint files, the LEFT HAND column refers to time locations in outfile of outdur duration, and the RIGHT HAND column refers to time locations in infile.ambitus – half-width of the region from within which the sound segments are read (seconds)
This is also true for the ambitus, step and clock parameters, except that the right hand column in these cases contains timing data, not times.
When used as a constant (single value), the locus time refers to a time location in the infile.
step – maximum length of (random) step between segment reads (> 0.002 seconds); this always falls within the ambitus: it is automatically adjusted where too large
clock – time between segment reads: this is the segment duration (> splicelen * 2) (seconds)
mindrnk – minimum number of clock ticks between sober plays (1 - 32767 Default: 10)
maxdrnk – maximum number of clock ticks between sober plays (1 - 32767 Default: 30)
-ssplicelen – length in milliseconds of the splice slope (Default: 15ms)
-cclokrand – randomisation of clock ticks (Range: 0 to 1 Default: 0)
-ooverlap – mutual overlap of segments in output (Range: 0 to 0.9900 Default: 0)
-rseed – any set value gives reproducible output
-llosober – minimum duration of sober plays (seconds) (Range: > 0 to duration of infile+. If >= duration of infile, all sober plays go to the end of the source.
-hhisober – maximum duration of sober plays (seconds)
(Range: > 0 to duration of infile+.All params except outdur, splicelen and seed may very in time.
Another approach to segmentation, EXTEND DRUNK takes a series of segments selected from the infile and splices them together to form the outfile. The process starts at some time in the file, called the locus and selects a segment, randomly, from within an 'ambit'. The length of 'ambit' is 2 * ambitus, and stretches to both sides of the locus position. Once a segment is read, the program moves (randomly) to a new position in either direction, within the ambit, and not more than step from the start location of the previous segment, from where it starts the next read. This is called a 'random walk'; hence the name 'drunk'.
The function is based on a drunken walk algorithm implemented by Miller Puckette.
While it is doing this walk, one can shift the locus, e.g., progressing slowly through the file. (NB: note above about times in the breakpoint file.) The 'ambit' – the portion of the soundfile being used at any one time, can be varied by altering the size of the ambitus, which is one-half the full 'ambit' width. The step, which is the maximum distance between the start of one read and the start of the next read (but must lie within the ambit), can also be varied.
For example, if the ambitus is small, segments very close to one another will be selected. Or if the step is much smaller than the segment size (a slow clock produces longer segments), selected segments will tend to overlap, producing random echoes or pre-echoes. The length of the segment, determined by the clock parameter is NOT constrained to the size of the ambit, so segments may begin within the ambit and end outside it. Segments which would end beyond the specified outdur are truncated.
To summarise, then, locus, ambitus and step all refer to start locations. Clock refers to segment length.
The process continues until outdur is filled, which makes it a useful program with which to generate material.
This is all about the fragmentation and texturing of sound. You can use this function to fragment a specific portion of a soundfile. Or you could make large jumps (step) in order to create surprise areas of fragmentation. Or EXTEND DRUNK can be used to churn up source material to varying degrees. The parameters allow a great deal of scope for variation, so some methodical study will be well rewarded. The following are some pointers regarding key parameters.
Musically, the issue is how much will the original soundfile be broken up. The various parameters contribute to this, each in their own way: the size of ambitus, the size of step, the length of the segment (clock and clokrand), and the position where this takes place (the locus). Introducing breakpoint files for time-varying effects adds another dimension again. Hence, for example, the locus might move gradually through the file, or moves back and forth in the file.
Clokrand randomises clock. If not used the output will consist of fixed lengths, a continuous stream of regular bursts of sound. Clokrand makes the segment lengths vary – in a time-varying way if a time clokrand breakpoint file is used.
Note that clock determines the length of the segments because the read continues until the next 'tick'. Overlap will increase the rate at which the segments come past: i.e., proportionately (0 to 0.9900) less than the length of each segment.
The splicelen (which can be quite long) can be used to soften the joins, thereby smoothing the output.
The seed parameter makes it possible to create a reproducible sequence. The 'random' sequence of numbers takes a fixed and hence repeatable form.
Finally, Mode 2 offers a hold mechanism whereby the unaltered infile continues to be read from where the read marker happens to be at the time. This can vary within limits, as set by mindrnk, maxdrnk, losober and hisober.
6 examples, 2-6 with breakpoint combinations
The infile used here is balsam.wav (1.997120 sec, mono, SR=44100), a vocal sound supplied with CDP's GrainMill.
To run these examples in Sound Loom, you should copy this sound (or a similar sound of the same length) into the Workspace directory (on the Workspace Page). You should also copy the files:
locus3.brk, locus4.brkonto the Workspace. You will find these files in the Support Pack drunkexamples.zip – NB: the command line .bat files are not needed for Sound Loom, which uses Patches. Then put balsam.wav on the Chosen Files list, press Process and select the EXTEND drunkwalk process.
ambitus1.brk, ambitus2.brk, ambitus3.brk, ambitus4.brk
step1.brk, step2.brk, step3.brk, step4.brk, step5.brk and
clock1.brk, clock2.brk, clock3.brk, clock4.brk, clock5.brkFor SoundShaper, make sure that that the sound is in your sounds directory, and that all the other files listed above are in your TXT directory. Then load the Presets (same names as the Patches) and run ProcessSF / Extend / Drunk. (Check the path to the .brk files to ensure that the Presets match your setup.)
For Command Line operation, put all files in the current directory and run the batch (.bat) files.
EXAMPLE 1 – fairly large segments are from widely spaced locations in the infile. Fixed values are used. Sound Loom/SoundShaper: Load the Patch/Preset drunk1.
INFILE OUTFILE LENGTH LOCUS AMBITUS STEP CLOCK extend drunk 1 balsam getdrnk1 25 1 .6 .2 .5EXAMPLE 2 – move gradually through the infile, with segment size (clock) decreasing, while the scattering (step) increases. Sound Loom/SoundShaper: Load the Patch/Preset drunk2.
LOCUS: left column is time in outfile, right column is time in infilelocus3.brk ambitus1.brk step1.brk clock1.brk 0 .3 0 .1 0 .05 0 .3 10 .6 10 .2 10 .1 10 .2 15 1.2 15 .3 15 .15 15 .1 20 1.8 20 .2 20 .2 20 .05 extend drunk 1 balsam getdrnk2 25 locus3.brk ambitus1.brk step1.brk clock1.brkEXAMPLE 3 – short segments (fast clock) are well-scattered within the full width (2 * ambitus) around each locus. Sound Loom/SoundShaper: Load the Patch/Preset drunk3.
locus3.brk ambitus2.brk step2.brk clock2.brk 0 .3 0 .2 0 .1 0 .05 10 .6 10 .2 10 .2 10 .1 15 1.2 15 .2 15 .3 15 .15 20 1.8 20 .2 20 .4 20 .1 extend drunk 1 balsam getdrnk3 25 locus3.brk ambitus2.brk step2.brk clock2.brkEXAMPLE 4 – longish segments (slow clock) which are located very close to each other (tiny steps) around each locus. Sound Loom/ SoundShaper: Load the Patch/Preset drunk4.
locus3.brk ambitus2.brk step3.brk clock3.brk 0 .3 0 .3 0 .05 0 .2 10 .6 10 .2 10 .08 10 .3 15 1.2 15 .3 15 .1 15 .2 20 1.8 20 .2 20 .06 20 .3 extend drunk 1 balsam getdrnk4 25 locus3.brk ambitus2.brk step3.brk clock3.brkEXAMPLE 5 – expansion outward from the centre: central locus, and the other parameters move from small to large. Sound Loom/SoundShaper: Load the Patch/Preset drunk5.
locus4.brk ambitus3.brk step4.brk clock4.brk 0 0.8 0 .1 0 .035 0 .05 8 1.2 10 .2 10 .09 10 .15 16 0.8 15 .3 15 .15 15 .2 20 .4 20 .2 20 .3 extend drunk 1 balsam getdrnk5 25 locus4.brk ambitus3.brk step4.brk clock4.brkEXAMPLE 6 – contraction inwards towards the centre: central locus, and the other parameters move from large to small. Sound Loom/SoundShaper: Load the Patch/Preset drunk6.
locus4.brk ambitus4.brk step5.brk clock5.brk 0 0.8 0 .4 0 .2 0 .3 8 1.2 10 .3 10 .15 10 .2 16 0.8 15 .2 15 .09 15 .15 20 .1 20 .035 20 .05 extend drunk 1 balsam getdrnk6 25 locus4.brk ambitus4.brk step5.brk clock5.brkMy thanks to Eitan Teomi whose queries helped me to understand DRUNK better and improve the documentation, and to Trevor Wishart for converting the command lines to Sound Loom Patches. [AE]
Also see DRUNK in the spectral set, which moves about in a similar way through analysis windows.
End of EXTEND DRUNK
Return to list of EXTEND functions at top
of this file
Return
to Main Index for the CDP System.
Return
to Groucho Index
1 Specify output duration
2 Specify number of repetitions
infile – input soundfile
outfile – output soundfile
outduration – Desired duration of resultant soundfile.
repetitions – Number of repetitions of frozen segment.
delay – The (average) delay between iterations: <= length of frozen segment.
rand – Delaytime randomisation. Range: 0 to 1. Default: 0.
pshift – Maximum of random pitchshift of each iteration. Range: 0 to 12 semitones. E.g., 2.5 = 2.5 semitones up or down.
ampcut – Maximum of random amplitude reduction on each iteration. Range: 0 to 1. Default: 0.
starttime_of_freeze – Time where the frozen segment begins in the original sound.
endtime – Time where the frozen segment ends in the original sound.
-sseed – The same seed number will produce identical output on rerun. Default: 0 – random sequence is different every time.
Extend a specific part of a sound using the iteration procedure. This tends to give a more convincing time-stretching result than any of the other time-stretch procedures, particularly as the non-time-stretched portions of the sound are not subject to any processing. The internal proportions of a sound event can be manipulated using this process.
The start and end times of the freeze enable you to focus on very specific parts of the sound, such as the 'a' in 'star' or the 's' in 'star'. With this program you can extend these to form sounds such as 'staaaaaaaaaaaaaaaaaaaar' or 'ssssssssssssssssssstar'. Given the claim that it gives 'a more convincing time-stretching result than any of the other time-stretch procedures', it is a program well worth exploring thoroughly.
End of EXTEND FREEZE
Return to List of functions for interacting
spectra at top of this file
Return
to Main Index for the CDP System.
Return
to Groucho Index
1 Iterate to a specified duration
2 Iterate a specified number of times
infile – input soundfile
outfile – output soundfile
outduration – length in seconds of outfile
repetitions – number of repetitions in the iteration
-ddelay – (average) delay between iterations in seconds (Default: length of infile)
-rrand – delay-time randomisation (Range: 0 to 1, Default: 0)
-ppshift – maximum random pitchshift of each iteration in semitones (Range: 0 to 12 semitones; e.g., 2.5 = 2.5 semitones up or down)
-aamp – maximum random amplitude reduction on each iteration (Range: 0 to 1, Default 0)
-ffade – (average) amplitude fade between iterations (Range: 0 to 1, Default 0)
-ggain – overall gain (Range: 0 to 1, Default: 0, which gives the best guess for no distortion)
-sseed – the same seed number will produce identical output on rerun (Default: 0 – the random sequence is different every time)
EXTEND ITERATE was written as a way of achieving more natural sounding iterations of a soundfile by introducing a randomisation of the delay time between each iterated segment, and slight variations in pitch or amplitude between the segments, as would occur in a naturally iterating source (e.g., a rolled 'rr' vocal sound). These randomisations can be selected (e.g., one might omit pitch variation, or not apply randomisation to the delay times), or applied in an exaggerated fashion, to achieve a number of different musical results.
The rand parameter introduces slight variations in delay between iterations, which may increase the 'naturalness' of the result. Omitting the rand parameter will produce a more mechanical echo effect.
The gain parameter allows some control over the amplitude of the mixed portions; the amount of gain suitable is dependent on the amplitude of the signal at the beginning and end of the soundfile (where the repeated units overlap). This can be examined with a soundfile viewer (such as VIEWSF, which can display the amplitude of each individual sample), and the gain adjusted accordingly if the defaults don't seem to be handling it properly. When randomisation is used, the gain is further reduced in the expectation that there will be a greater degree of overlap.
This function produces a series of (usually overlapping) repeats of a soundfile. The nature of the attack portion of the soundfile – sharp or gradual – will greatly affect the way these repetitions are perceived. The use of a very short soundfile, e.g., 0.2 seconds, especially one with a sharp attack, will result in a rapid-fire succession of easily perceived iterations.
End of EXTEND ITERATE
Return to list of EXTEND functions at top
of this file
Return
to Main Index for the CDP System.
Return
to Groucho Index
1 Segment advances in soundfile until soundfile is exhausted
2 Specify outfile duration (shortened if looping reaches end of infile)
3 Specify number of loop repeats (reduced if looping reaches end of infile)
infile – soundfile to process
outfile – output soundfile
dur – duration of outfile required (in seconds)
cnt – number of loop repeats required
start – time (in seconds) in infile at which the looping process begins
len – length of looped segment (in milliseconds)
[-l]step – advance in infile from the start of one loop to the next (in milliseconds)May be zero in Modes 2 and 3 but not in Mode 1. When zero, repeating loops of the same material are created.-wsplen – length of splice (in milliseconds) (Default: 25ms)
-sscat – make step advance irregularly, within the timeframe given by scat
-b – play from beginning of infile (even if looping doesn't begin there)
The key feature of this process is that it joins together, end-to-end, a series of segments taken from the file, each with a splice slope to avoid clicks. These segments are all of the same length, so one way or another, the result may appear to have some degree of regular pulsation. This does not (usually) result from the presence of splices, but rather is a perceptual result caused by the repetition of sonic material.
The most salient parameters are step and len. Step is the timestep in the sourcefile between the start of one selected segment and the next.
- If the step is zero, the selected segment will simply be repeated.
- If the step is > 0 but smaller than the segment length, the selected segments will share much in common, but each will begin at a point in the sourcefile a little later than the last, producing a progressing echo effect. The outfile will be longer than the infile.
Diagram 1
- If the step is larger than the segment length, the process will leap through the file, omitting bits of it, and rapidly get to the end. The result will be shorter than the infile.
Diagram 2
The scat parameter randomises the length of step (within a small range), producing a less mechanical result.
The length of the segments (len) – as well as the size of the step – affects the recognisability factor of the original source material.
The following notes on the subject of 'recognisability' are a paraphrase, and any inaccuracies in the reporting are due to me (Ed.).
In his talk at the Sonic Arts Network Conference in Huddersfield on 14 March 1999, featured composer Yves Daoust said that he always sought to respect and usually be able to recognise the source when concrète materials are used. His point was that if recognisability is removed completely, the sound no longer has a meaning: i.e., you could have produced that result in the studio using any type of source material. So the issue of why you are using certain types of source material is integral to the whole conception of the piece – and, he emphasised, it is very important to come to the studio with a strong conception; otherwise you could be overly influenced by the ramifications of the source material and not, as a result, achieve a clear and focused musical form.He found he was working towards 'a more abstract way' to deal with the musical structures. He described the beginnings of several such ideas:
- fragmentation of material (which is not the same as 'granulation')
- a 'scale' between 'unbroken sound' (no fragmentation) and 'granulation' (complete dispersal of sonic material)
- the filtering of information (rather than just filtering in a sonic sense)
- morphological filters
- sieves: small apertures in the e.g. frequency spectrum plus 'enough energy to get out'
- a common rule applied to disparate material
- trajectories: large scale shaping, such as long filter sweeps, slowly changing density, long glissandi (realised in different ways)
Musical applications
EXTEND LOOP can be used for pulsating a sound in a regular way. The incremental movement through the source can be illustrated by moving through the word 'anchovies' with a step that moves ahead one letter at a time and a segment length which encompases 4 letters:
Because of the limited nature of these facilities, EXTEND LOOP can be used to play with sounds in a controlled way. Some of the effects resulting from very small steps and segment lengths will be surprising and approach brassage techniques.anch-ncho-chov-hovi-ovie-viesFor (much) more flexibility in brassage and granulation, see MODIFY BRASSAGE and its graphic counterpart GrainMill (on Windows 95 sytems). Other functions which carry out fragmentation in some way include EXTEND DRUNK, EXTEND SCRAMBLE and EXTEND ZIGZAG (in this file).
In the spectral domain, also see BLUR DRUNK, BLUR SHUFFLE, and BLUR WEAVE. COMBINE LEAF fragments by interpolating N analysis windows from two different files, functions such as FOCUS FREEZE and FOCUS STEP fragment by holding material according to a time frame pattern. MORPH BRIDGE and MORPH GLIDE break up a file by working with specified analysis windows, and SPEC GRAB and SPEC MAGNIFY can isolate and extend a single analysis window. Finally, the whole DISTORT package fragments material by creating pseudo-wavecycles from sonic material found between zero crossings.
End of EXTEND LOOP
Return to list of EXTEND functions at top of this file
Return to Main Index for the CDP System.
Return to Groucho Index
EXTEND REPETITIONS – Repeat source at given times
Usage
extend repetitions infile outfile timesfile level
Parameters
infile – input soundfile
outfile – output soundfile
timesfile – Textfile of times (in seconds) at which the source plays.
level – Level of output. Range: 0 to 1.
Level may vary over time.Understanding the EXTEND REPETITIONS Function
This program can be thought of either as a more controlled looping function or a simple rhythm sequencer.
Musical applications
Controlled looping or rhythmic sequencing.
End of EXTEND REPETITIONS
Return to List of functions for interacting spectra at top of this file
Return to Main Index for the CDP System.
Return to Groucho Index
EXTEND SCRAMBLE – Scramble soundfile and write to any given length
Usage
extend scramble 1 infile outfile minseglen maxseglen outdur [-wsplen] [-sseed] [-b] [-e]
extend scramble 2 infile outfile seglen scatter outdur [-wsplen] [-sseed] [-b] [-e]
Modes
1 Cut random chunks from infile and splice end to end
2 Cut infile into random chunks and rearrange; repeat differently, etc.
Parameters
infile – input soundfile to process
outfile – output soundfile
minseglen – minimum chunksize to cut
maxseglen – maximum chunksize to cut (Range: 0.045 to length of infile – must be > minseglen)
seglen – average chunksize to cut
scatter – randomisation of chunk lengths (>= 0)Cannot be greater than infilesize/seglen (rounded down)outdur – duration of outfile required (> maxseglen
-wsplen – duration of splice in milliseconds Default: 25ms)
-sseed – the same seed number will produce identical output on rerun (Default: 0, random sequence is different every time)
-b – force start of outfile to be beginning of infile
-e – force end of outfile to be end of infile
Understanding the SCRAMBLE Process
With EXTEND SCRAMBLE, segments of soundfile are selected from a wide variety of locations in the infile, jumping back and forth a great deal.
Mode 1 takes the infile, chooses a random chunk of it, and then chooses another random chunk of it which may overlap with the first choice, then another chunk which may overlap with either of the other two ... etc. Then it splices them all together. Thus, any bits of the file may be repeated quite quickly if overlapping material is selected in consecutive chunks, and some bits may not appear at all if never randomly selected.
The size of the chunks will be a random length somewhere between minseglen and maxseglen.
Mode 2, cuts the entire file into random-length chunks which do not overlap. It arranges these at random. The process is then repeated, but the random cuts are of course in different positions in the file. Consequently, the entire file is used, and used only once, before the process starts to use the file again.
In Mode 2 an average chunksize is specified plus a random factor (scatter). The formula which shows what the maximum scatter factor can be reveals that Mode 2 can be used to make chunks which vary a great deal in length. For example, if the infile is 2 seconds long and seglen is 0.3, the maximum value for scatter will be 6.0 (rounded down). (This value was accepted – and worked – even with an outdur of 4.0.)
The 2nd Mode also provides the option to rerun with identical output.
The ability to write to any length of outfile makes it possible to give the process plenty of time to make full use of the infile.
Musical applications
EXTEND SCRAMBLE provides a relatively automated way to fragment a soundfile in a random way, tending to swing back and forth from the beginning and end portions of the infile. The jumping about is likely to be extreme, so applied to vocal material, the results will be somewhere between wild and funny. Applied to pitched material, the result can sound like an improvisation.
For a more carefully defined zigzagging motion through a soundfile, see EXTEND ZIGZAG (below).
End of EXTEND SCRAMBLE
Return to list of EXTEND functions at top of this file
Return to Main Index for the CDP System.
Return to Groucho Index
EXTEND SEQUENCE – Produce a sequence from an input sound played at specified transpositions and times
Usage
extend sequence infile outfile sequence-file attenuation
Parameters
infile – input soundfile
outfile – output soundfile, being a sequence made from the (one) input soundfile, according to the instructions in the sequence-file.
sequence-file – contains 3 values on each line, separated by tabs or spaces, one line per event: output-time semitone-transposition loudness value triples, where loudness is a loudness multiplier. There needs to be one value-triple for each event in the sequence.
attenuation – overall attenuation to apply to the source, should outfile overloadUnderstanding the EXTEND SEQUENCE Function
This program works like a simple conventional sequencer except that it takes only one input (see EXTEND SEQUENCE2 for multiple soundfile input. The process takes a sequence-file of triple-values:
for each event in the sequence. The source sound is then copied at each output-time, transposed by each transposition) amount in (fractional) semitones, and attenuated to the level specified. The result is a sequence of events derived from the one source sample.
- output-time – the time when you want the soundfile to come in again in the outfile
- transposition – the pitch-level of that entry, given in (possibly fractional) semitones
- amplitude – the relative level of that sound in the output sequence, louder (> 1.0) or softer (< 1.0)
Musical applications
This function was used, for example, to make the underlying sequence of the 'Gamelan' in my (T Wishart) composition, Imago. TW: in Trevor Wishart's compositon, Imago. Transposition or time sequences might be derived from data from other sounds, generated in the Sound Loom Table Editor or with COLUMNS, or entered by hand in a text file. EXTEND SEQUENCE is therefore a useful way to create rhythmic textures, whether simple or very intricate.
End of EXTEND SEQUENCE
Return to List of functions for interacting spectra at top of this file
Return to Main Index for the CDP System.
Return to Groucho Index
EXTEND SEQUENCE2 – Produce a sequence from several sounds played at specified transpositions and times
Usage
extend sequence2 infile1 [infile2 ...] outfile sequence-file attenuation
Parameters
infile1 – input soundfile
[infile2] – optional additional input soundfiles. All input files must have the same number of channels.
outfile – resultant output sequence of soundfiles
sequence-file – data file in which the first line contains notional MIDI pitch values for each input soundfile and each subsequent line contains 5 values on each line, given in order from left to right, separated by spaces:attenuation – overall attenuation to apply to the source, should outfile overload
- input-sound-number – these numbers follow the order in which they are given as inputs
- output-time – time in the outfile when this event is to begin
- MIDI pitch – MIDI pitch level at which to perform the sound, relative to the notional pitch given in Line 1 – it may be fractional, i.e., microtonal, such as 60.5, which is ½ semitone (50 cents) higher than Middle C.
- loudness – relative loudness multiplier for that event. Range: 0 to 1.
- duration – a duration for that event: it can curtail it (i.e., truncate the source sound), but cannot extend it. Note that transposition needs to be taken into account: the maximum event length = the (transposed) duration of the sound chosen. (The transposed length will be the source length * the transposition ratio – the Music Calculator converts between transposition in (fractional) MIDI pitches and transposition in ratios.)
Understanding the EXTEND SEQUENCE2 Function
An example sequence-file serves as a reminder of how the data is put together:
60 60 [Snds Stime MPV Level Dur] 1 0.0 60 0.25 1 2 0.5 62 0.50 2 1 1.5 63 0.25 1.5 2 2.5 64 0.50 2Here we have a convenient way to arrange several different files in a rhythmic way, with several additional parameter settings.
- There are two input sounds. The first one given to the program will be No. 1 and the second one will be No. 2.
- The Start times set the times at which they begin to play.
- The MPVs (MIDI Pitch Values) specify the pitch level relative to the notional MPVs given in the first line, one for each file, which may or may not be the actual pitch level of the sound.
- Then the amplitude level is specified, so that certain sounds can be emphasised, different original amplitudes readjusted, etc.
- Finally, the duration of the note event is given. It may be shorter than the original sound, but, of course, not longer. This provides an easy way to work with the attack portion of a sound, or to layer longer sounds.
Musical applications
We are familiar with standard MIX files. They specify the sound by using its name. Here we specify the sound by using a number. This means that it is easy to list and order the sounds, as well as to use a numerical pattern generated in some other way, such as algorithmically. Also different from the standard MIX files are the pitch transposition and duration fields.
If the sound is fairly clearly pitched, such as a bell sound, then EXTEND SEQUENCE2 enables us quickly to:
Note that the same sound can be repeated. The functionality of the program is shown by combining its use with the table editing software in the CDP System. Thus we can massage the columnar data with the Table Editor in Sound Loom (= DATA, Columns in Soundshaper or just columns on the Command Line). Several different versions could be made and then each one realised with EXTEND SEQUENCE2 by loading in the various sequence-files in turn. For example, a structural ritardando could be made by adding a value to the start_time column and subtracting a value from the pitch_level column 3 or 4 times. The result is a series of output soundfiles in which several sounds repeat, placed further and further apart in time, while getting closer together in pitch.
- Create a melody
- Create a lively rhythmic pattern, with the same or different pitches
- Pattern the loudness in a sequence of sounds
- Use the attack of the sound we want while discarding the remainder
- Create some 'changes' as used by bell-ringers – or any other intricate sequence of notes
End of EXTEND SEQUENCE2
Return to List of functions for interacting spectra at top of this file
Return to Main Index for the CDP System.
Return to Groucho Index
EXTEND ZIGZAG – Read soundfile backwards and forwards, as you specify
Usage
extend zigzag 1 infile outfile start end dur minzig [-ssplicelen] [-mmaxzig] [-rseed]
OR:
extend zigzag 2 infile outfile timefile [-ssplicelen]
Modes
1 Random zigzags: start and end at beginning and end of infile
2 Zigzagging follows times supplied by the user
Parameters
infile – input soundfile to process
outfile – output soundfile
start – together with ...
end – define the time interval in which the zigzagging occurs
dur – total duration of output sound required
minzig – minimum acceptable time between successive zigzag timepoints
-ssplicelen – length of splice slope in milliseconds (Default: 25ms)
-mmaxzig – maximum acceptable time between successive zigzag timepoints
-rseed – number to generate a replicable random sequence (> 0 – Default: 0, random sequence is different every time)Entering the same number on the next program run generates the same sequence.timefile – text file containing sequence of times to zigzag betweenEach step-between-times must be > (3 * splicelen). NB: Zigsteps moving in the same (time-)direction will be concatenated.Understanding the EXTEND ZIGZAG Process
What is special about this process is that it actually reads the source soundfile backwards when it moves from a later to an earlier timepoint.
It is expected that a user timefile will normally alternate between earlier and later times. Moving in the same direction is a bit pointless. If two steps move in the same direction, a Warning message is generated, on the assumption that this was an accidental entry by the user, e.g., a time was omitted. However, it still produces an output.
The timefile is written as a series of times in seconds, either horizontally with spaces between the times, or with each time on a separate line.
Musical applications
Because moving from later to earlier times reads the source backwards, the effect of the zigzag transforms the sonic material much more than with EXTEND SCRAMBLE. The degree to which this happens depends on the duration between earlier and later times when moving from earlier to later, i.e., how much of the source is actually read forwards.
EXTEND ZIGZAG is a good way to begin to transmute sonic material into something more abstract. It can also be handled in such a way as to create a powerful warping effect: by making the forwards and backwards reads over fairly long durations. Short backwards and forwards movements create a stuttering or repeated note effect. (See diagram below, by Louisa Yong).
Diagram of a Zigzag
End of EXTEND ZIGZAG
Return to list of EXTEND functions at top of this file
Return to Main Index for the CDP System.
Return to Groucho Index
Last Updated: 27 February 2005
Documentation: Archer Endrich
All observations & ideas for improvement appreciated
Composers Desktop Project
12 Goodwood Way
Cepen Park South
Chippenham, Wiltshire
SN14 0SY ENGLAND
Tel. +44 (0)1249 461361
Email: archer@trans4um.demon.co.uk
© Copyright 1998-2005 Archer Endrich & CDP