The main granular program in CDP is MODIFY BRASSAGE.
The GRAIN functions here are for manipulating grains in a "grainy" sound, normally one with gaps.
Note that a sound can be quickly granulated with MODIFY BRASSAGE Mode 5 with a density value of 1.0 or less, which will introduce gaps between the grains:
&bsp;
modify brassage 5 infile outfile density
Alternatively, useENVNU PEAKCHOP combined with ENVEL REPLACE.
Names in brackets mean that these are separate programs. The others are functions within GRAIN.
infile1 – 1st input soundfile: provides grain onset times
infile2 – 2nd input soundfile: provides the actual grains to be re-timed
outfile – output soundfile
offset – add this value to all grain timings
gate2 – minimum signal level to register grain in infile2 (Range: 0 to 1)
-blen – maximum time between grains (Range: 1 to duration of infile)
-lgate – minimum signal level to register grain in infile1 (Range: 0 to 1, Default 0.3)
-hminhole – minimum duration of holes between grains (Minimum allowed value is 0.032 – the Default)
-twinsize – gate level tracks the signal level, as found with a window size of winsize milliseconds (Range: 0.0 to duration of infile1)0.0 turns off tracking-x – ignore the last grain in the source
gate and gate2 may vary over time.
The timing of the grains of infile2 is made to match those of infile1.
It may be unwise not to specify a value for the optional parameter -lgate (gate level for infile1).
If you achieve an effective grain timing for two sounds, GRAIN ALIGN makes it possible to create a subtle correspondence between the two soundfiles.
For example, the fine attack-structure of the two sounds can be synchronised, so that they appear to pulse in parallel.
End of GRAIN ALIGN
Return to list of GRAIN functions at top
of this file
Return
to Main Index for the CDP System.
Return
to 'Groucho' Time-Domain Index
infile – input soundfile in which to estimate a useful gate value
GRAIN ASSESS is a simple utility to provide useful information when using the GRAIN set of programs. You just provide an input soundfile and it scans the file, displaying a report such as:
Maximum grains found = 21 at gate value 0.117100 and windowlen 50ms
The program is therefore assessing which gate value and which windowlength value will result in the most grains. You can then use this information in other GRAIN programs.
For example, even as simple a program as GRAIN COUNT gives you the option to specify a gate value. GRAIN ASSESS provides the value that will produce the most grains, and you can take it from there. Windowlength also affects the number of grains found. If it is too long, it may count several grains as one, or if too short, it may not find some of the grains.
End of GRAIN ASSESS
Return to list of GRAIN functions at top
of this file
Return
to Main Index for the CDP System.
Return
to 'Groucho' Time-Domain Index
infile – input soundfile in which to count the grains
-blen – maximum time between grains (Range: 1 to duration of infile)
-lgate – required signal level for grain to be seen (Range: 0 to 1; the Default is 0.3)gate may vary over time.-hminhole – minimum duration of holes between grains (Minimum allowed value is 0.032 – the Default)
-twinsize – gate level tracks the signal level, as found with a window size of winsize milliseconds (Range: 0.0 to duration of infile)0.0 turns off tracking-x – ignore the last grain in the source
GRAIN COUNT displays the message: "N grains found at this gate level."
GRAIN COUNT provides a check on the granularity of a file, and also on the gate level at which a given number of grains is operational.
End of GRAIN COUNT
Return to list of GRAIN functions at top
of this file
Return
to Main Index for the CDP System.
Return
to 'Groucho' Time-Domain Index
infile – input soundfile
N – number of repetitions of each grain
-blen – maximum time between grains. (Range: 1 to duration of infile)
-lgate – required signal level for grain to be seen (Range: 0 to 1; the Default is 0.3)N and gate may vary over time.-hminhole – minimum duration of holes between grains (Minimum allowed value is 0.032 – the Default)
-twinsize – gate level tracks the signal level, as found with a window size of winsize milliseconds (Range: 0.0 to duration of infile)0.0 turns off tracking-x – ignore the last grain in the source
This process makes N copies of each grain, proceding through the infile grain by grain.
You are recommended to use GRAIN COUNT with the gate flag to determine which gate level works best with the file (produces a good granulation of the source). This gate level should be used with GRAIN DUPLICATE. Not using it appears to produce anomalous results, such as no grain duplications and a long period of silence.
The overall effect is that of rapid-fire repeats or a stuttering effect, depending on the source.
End of GRAIN DUPLICATE
Return to list of GRAIN functions at top
of this file
Return
to Main Index for the CDP System.
Return
to 'Groucho' Time-Domain Index
infile – input soundfile in which to find the grain-onsets
out-textfile – output textfile containing a list of times in seconds at which each grain begins
-blen – maximum time between grains (Range: 1 to duration of infile)
-lgate – required signal level for grain to be seen (Range: 0 to 1; the Default is 0.3)gate may vary over time.-hminhole – minimum duration of holes between grains (Minimum allowed value is 0.032 – the Default)
-twinsize – gate level tracks the signal level, as found with a window size of winsize milliseconds (Range: 0.0 to duration of infile)0.0 turns off tracking-x – ignore the last grain in the source
This function locates and writes to a text file the time in seconds at which each grain begins. Again, the effective gate level determined with GRAIN COUNT should be used.
Examination of this text file provides a way to check on the results of the various GRAIN functions which affect the grain timings. The data can be used to retime the grains in another grainy sound. See GRAIN REPOSITION.
End of GRAIN FIND
Return to list of GRAIN functions at top
of this file
Return
to Main Index for the CDP System.
Return
to 'Groucho' Time-Domain Index
Example command line to find and extend grain regions:
grainex extend in.wav out.wav 30 0.5 2.5 1.5 3.5
OR:WSIZ TROF PLUS STT ENDgrainex extend inf outf 5 0.5 5 0 1
infile – input soundfile in which to find the grains
outfile – output soundfile
wsiz – size of window in milliseconds, which determines the size of the grains to find (Range: 1.81406 to file-length ÷ 3ms)
trof – acceptable trough depth, relative to adjacent peaks (Range: > 0 to < 1 – Soundshaper default: 0.5 – see Note below)
plus – how much duration to add to the source (Range: 0.000002 to 3600 secs)
stt – time of start of grain material within the source (Range: 0.0 to file-length in seconds)
end – time of end of grain material within the source (Range: 0 to file-length in seconds and > stt)
For GRAINEX, as with the other GRAIN functions, you will need a reasonably grainy sound as an input, otherwise it won't be able to find any grains and you will probably see the message "Insufficient valid troughs in the file".
GRAINEX extends an area containing grains, the start and end times being set by the user. The grains are found by envelope troughs and zero-crossings.
When the above example command line was run with the input file count.wav, when the output sound got to the number 'three', it was repeated several times before continuing on with the rest of the count up to the number ten. The input sound was 8.066 seconds long, and the output was 18.866 seconds long, i.e., considerably more than the 2.5 seconds specified.
NOTE
It is possible to set trof too low. Being relative to the peak level, if the signal doesn't drop that far in level, you may get the Error Message: "Insufficient grains to proceed" or "INSUFFICIENT PEAKS IN THE FILE AREA SPECIFIED".
...
End of GRAINEX EXTEND
Return to list of GRAIN functions at top
of this file
Return
to Main Index for the CDP System.
Return
to 'Groucho' Time-Domain Index
Example command line:
grain grev 2 infile outfile 10 0.5 4 3
1 REVERSE – the grain-units isolated are played in reverse order
2 REPEAT – gpcnt-sized units are repeated
3 DELETE – remove the specified number of units
4 OMIT – replace specified number of units with silence
5 TIMESTRETCH – the time expansion is happening in the troughs: the grains of sound are not themselves stretched
6 GET – writes grain time-positions to a textfile
7 PUT – places grains in time according to textfile
infile – input soundfile in which to count the grains
outfile – output soundfile produced by the program
out_timesfile – textfile produced by the program, containing times of grain positions
in_timesfile – textfile containg times at which the program is to place grains
wsiz – windowsize in milliseconds, determining the size of the grains to find
trof – the acceptable trough height, relative to adjacent peaks. Range: greater than 0 to less than 1 (> 0 Range < 1)
gpcnt – ('groupcount') the number of grains to treat as a unit in the operations
repets – the number of repetitions of each unit
keep – outof – the number of units to keep, e.g., 3 out of 5
tstretch – the amount to timestretch the output (grains NOT stretched) Range: 0.01 to 100)gpcnt, repets, keep and tstretch can vary over time.
This program was written to assist in separating the large-scale 'grains' in a stream of sound, e.g. the syllables of speech. It can be a useful alternative to SFEDIT SYLLABLES, which is for the precise manual editing of syllabic units.
The other 'grain' programs rely on a gate which searches for moments when the input sound falls below a certain level, and then splices the sound at those points. This works fine when the sound grains are fairly consistent (e.g. a sequence of pizzicato sounds).
However, the syllables of speech are usually quite different from one another. So the gating procedure finds some good places to cut, and misses others. The 'grev' process does not depend on setting a gate level. It simply looks for troughs (low points) in the envelope (at a suitable timescale), then within those troughs, searches for a zero-crossing at the lowest point of the signal in the trough.
Once it finds these it can separate the grains with zero-length splices. The key thing is to set the envelope window about three times smaller than the features (e.g. syllables) you want to tease out.
Designed for spoken words, this program enables you to play with syllabic units with remarkable efficiency. Although similar to others in the GRAIN set, it gives direct control over the syllabic content of speech.
Mode 2 reveals how this program works. The example command line given above has these parameters: wsiz = 10 ms, trof = 0.5 (half-way can be a useful place to start with a parameter), gpcnt = 4 (four grains to a group), and repets = 3 (repeat the 4-grain-units 3 times). When run, we find that the syllables of the spoken text are, for the most part, clearly isolated and repeat 3 times each without overlaps. Thus the program is working effectively with the spaces between the syllables.
That this is the case is also shown by Mode 5. When run with the same parameter values (the last parameter being tstretch instead of repets, we find that the syllables do not repeat but are more spaced out in time.
Modes 3 and 4 operate as expected, deleting grains or replacing them with silence. Mode 1 plays back the isolated grains in reverse order, i.e., you hear syllabic fragments from the end of the file first and proceed back to the beginning – but the text fragments themselves are 'forward' as normal. Higher values for wsiz and gpcnt result in longer passages of the text treated as a unit, while smaller values fragment the text more. It's a very odd effect.
Mode 6 provides a way to write grain-position times (shaped by the other parameters) to a file. This file can then be used as input in Mode 7 with, e.g., another input sound or with different values for the other parameters, thus achieving results that are partly shaped and partly serendipitous.
Altogether, GRAIN GREV is a powerful tool with which to manipulate speech material. On a deeper level, it can be used to musicalise speech and make formal connections with other shapes in your composition.
End of GRAIN GREV
Return to list of GRAIN functions at top
of this file
Return
to Main Index for the CDP System.
Return
to 'Groucho' Time-Domain Index
Example command line to create a combo grain texture :
newtex newtex 3 insndfile.wav outsndfile.wav 30 4 3 0.1 0 0 1.5 0.35 -s36
i.e.:DUR CHANS MAXRANGE STEP SPACETYPE LOC AMB GSTEP SPLICE 30 4 3 0.1 0 0 1.5 0.35 36This example takes short segments and scatters them randomly across the channels. Splice is quite long to give the segments a softer attack. It works well in stereo too, but 4 or more channels are worth generating to get a variety of segmentations. (The channels can be split, possibly processed and then re-combined or re-mixed in various ways.)
1 The transpositions of insndfile are spread over N octaves and spatially, and fade in and out randomly.
2 Insndfile is read at its original rate (i.e., no transpositions), spread spatially, and fades in and out randomly.
3 Insndfile is read as 'drunken walks', spread spatially, and fades in and out randomly.
insndfile(s) – input soundfile or soundfiles, depending on which Mode is used
outsndfile – the output textured soundfile
transposes – textfile containing a list of transposition ratios and relative levels, against time.
dur – the duration of the output sound
- The data is a text file of lines of data.
- Every line must have the same number of entries.
- The first entry on each line is a time.
- Times must start at zero and increase.
- All EVEN-numbered entries are transposition levels.
- Transpositions must increase from entry to entry.
- All other ODD-numbered entries are transposition levels.
- Levels should have values between -1 and 1.
- Octave (transposition) values invert the phase of the source.
chans – the number of output channels
maxrange – In Mode 1 maxrange is the range in octaves of the transpositions of the input.
In Mode 2 maxrange is the number of simultaneous soundings of any source soundfile.
step – the average time between changes to the stream-content of the output.
splice – the splice-lengths in milliseconds for component entry and exit
number – the number of components chosen for each event
delay – the time delay between identical components
loc – the locus (time-location) from which to read sound segments.
NB: It is an important feature that loc is time-varying.
amb – the 'ambitus', i.e., the restricted time-range in milliseconds around the locus where reads can begin.
NB: It is an important feature that amb is time-varying.
• TIP: Compare locus and ambitus with the parameters of the same name in EXTEND DRUNK.
gstep – the maximum size of the random steps between one read start and the next read.
-x – 'Xclusive': change all components, as far as possible, from event to event
-j – 'Jump': all components are assigned to the same location for any one event, and then it jumps to the next location
spacetype – the type of output spatialisation:
0 Spatial position set at random.
For 8-channel output only (assumes an 8-speaker 'ring' formation):EXTRA OPTIONAL FLAGS (only if spacetype>0):
- Positions alternate between Left and Right sides, but are otherwise random.
- Positions alternate between Front and Back, but are otherwise random.
- Rotating clockwise or anticlockwise.
- Random permutations of all 8 channels.
- As [4] plus all possible pairs of channels.
- As [4] plus all possible meaningful small and large triangles.
- As [4] plus square, diamond and all-at-once.
In types 4 to 7, all members of the permutation are used before the next permutation starts.- Alternate between all-left and all-right.
- Alternate between all-front and all-back.
- Alternate between all-square and all-diamond.
- Rotate triangle formed by loudspeakers 2-apart clockwise.
- Rotate triangle formed by loudspeakers 3-apart clockwise.
- Rotate triangle formed by loudspeakers 2-apart anticlockwise.
- Rotate triangle formed by loudspeakers 3-apart anticlockwise.
rotspeed – rotation speed for certain spatialisation types
-efrom -Etime– 'Emerge': the sound emerges from channel from over time time at start
-cfrom -Ctime– 'Converge': the sound converges from channel to over time time at end
- Flags with NO parameters must be placed AFTER any flags WITH parameters, on the command line.
- maxrange, step, number, loc, amb and gstep, can vary over time.
NEWTEX is a powerhouse of a program that combines aspects of MODIFY BRASSAGE, EXTEND DRUNK and TEXTURE. It generates a (time-varying) texture from segments cut from a source sound or sounds.
...
End of NEWTEX NEWTEX
Return to list of GRAIN functions at
top of this file
Return to Main Index for
the CDP System.
Return
to 'Groucho' Time-Domain Index
infile – input soundfile in which to count the grains
outfile – output soundfile produced by the program
duration – duration of the noise part of the output soundfile
minfrq – the lowest 'frequency' in Hz acceptable as noise. Range: 1000 to 22050Hz. (Try 6000Hz)
mindur – the minimum duration of signal in milliseconds acceptable as noise source. Range: 0 to 50 milliseconds
maxdur – the maximum duration of signal in seconds acceptable as noise source. Range: 0.0 to length of input soundfile (in seconds)
-x – keep only the extended noise. The Default is to keep the rest of the input source sound as well.
This process is complementary to GRAIN R_EXTEND. It searches for sybillants in speech (or noise materials in any input sound) then allows them (and only them) to be sustained.
Note that different types of values are involved with the mindur and maxdur parameters. The minimum duration is given in milliseconds, but the maximum duration, because you're thinking in terms of the length of the soundfile, is given in seconds.
The -x option enables you to save only the extended noise component, discarding the rest of the file. When -x is not invoked, you hear the whole soundfile (up to the maxdur you specified) with that first (extended) noise component in its original position in the soundfile.
This program acually looks for and extends (only) the first noise component that it finds at the given parameter specifications.
The extended noise component, especially if saved on its own via the -x option, can be a useful bit of source material for other sound transformations.
End of GRAIN NOISE_EXTEND
Return to list of GRAIN functions at top
of this file
Return
to Main Index for the CDP System.
Return
to 'Groucho' Time-Domain Index
infile – input soundfile to process
outfile – output soundfile
keep – number of grains to keep from each set of out-of grainskeep may vary over time, but must not exceed out-ofout-of – keep grains retained from start of each set of out-of grains
-blen – maximum time between grains (Range: 1 to duration of infile)
-lgate – required signal level for grain to be seen (Range: 0 to 1; the Default is 0.3)gate may vary over time.-hminhole – minimum duration of holes between grains (Minimum allowed value is 0.032 – the Default)
-twinsize – gate level tracks the signal level, as found with a window size of winsize milliseconds (Range: 0.0 to duration of infile)0.0 turns off tracking-x – ignore the last grain in the source
The key here is the out-of parameter. This defines the size of the group which serves as the unit of operations: out of every unit, keep grains are retained. Thus, if 5 out of 10 are retained, 50% of the file is omitted, if 7 out of 10 are retained, 3 out of the 10 are omitted (30%go), if 4 out of 10 are retained, 6 out of the 10 are omitted (60% go).
What is significant about the out-of parameter is that its size affects the result, Thus, a large value for out-of with a smaller value for keep , e.g., 20 and 10, will break up the file much more than it would if out-of were smaller: e.g., 4 and 2 – even though the proportion of omitted material is the same (50%). This is because larger continuous chunks of the source (10 grains rather than 2 grains) are lost in the former case.
GRAIN OMIT provides a useful way to contract sound material without altering the sonic substance (i.e. the grains are not themselves, time-contracted).
End of GRAIN OMIT
Return to list of GRAIN functions at top
of this file
Return
to Main Index for the CDP System.
Return
to 'Groucho' Time-Domain Index
1 Transform each grain in turn, without repeating any grains: on reaching the end of the transpmultfile data, cycle back to its start.
2 Transform grain in each specified way before proceeding to the next grain.
infile – input soundfile to process
outfile – output soundfile
transpmultfile – is a file containing transposition time multiplier pairs.-blen – maximum time between grains (Range: 1 to duration of infile)
- transpositions are given as positive or negative semitone shifts.
- the maximum transposition is 4 octaves up or down
- time multipliers change the duration between one grain onset and the next grain onset.
- the maximum value for a multiplier is 1000
- the minimum value for a multiplier is 0.001000
- If any inter-grain time is reduced the minimum time allowed for a grain (0.032 sec), it will be reset to this minimum grain time.
-lgate – required signal level for grain to be seen (Range: 0 to 1; the Default is 0.3)gate may vary over time.-hminhole – minimum duration of holes between grains (Minimum allowed value is 0.032 – the Default)
-twinsize – gate level tracks the signal level, as found with a window size of winsize milliseconds (Range: 0.0 to duration of infile)0.0 turns off tracking-x – ignore the last grain in the source
Implementing this process is really very easy. The transpmultfile does not have to match the number of grains in the infile. You just need to design the pitch and time shapes into which you want the grains to flow. The pitch transpositions are placed in the left column as positive or negative semitones, and the time-multipliers are placed in the right column, > 1 to make the gap longer, < 1 to make it shorter.
Here is an example transpmultfile that rises, drops down below the original pitch level, and then rises back to the starting point, with the grains getting shorter and then longer again.
[transposition timing] (semitones) (multiplier) 0 1.0 1 0.9 2 0.8 3 0.7 2 0.6 1 0.5 0 0.4 -1 0.5 -2 0.6 -3 0.7 -2 0.8 -1 0.9 0 1.0The difference between the modes is dramatic.
- In Mode 1 a single line of the transpmultfile is applied to a single grain, the next line to the next grain etc., cycling around the textfile until reaching the last grain of infile. Any difference in duration between infile and outfile is due to the timing difference between the grains. You can make the shape occur once by having as many lines in the textfile as there are grains in the infile.
- In Mode 2 ALL of the lines of the transpmultfile are applied to each grain in turn. Thus every grain repeats for as many times as there are lines in the textfile, and one hears the whole shape defined in the textfile repeated as many times as there are grains in the infile. This makes the outfile much longer than the infile.
Be careful with Mode 2, because an over-long textfile will create a very long outfile (and be very repetitive).
GRAIN REMOTIF can be used to apply a gross shape to the grains of the infile, with controlled accelerandi or other timing changes. This shape can occur once, cycle around the textfile with one line per grain (Mode 1), or the shape can occur once per grain: each grain repeats with transposition and timing differences as many times as there are lines in the textfile (Mode 2).
End of GRAIN REMOTIF
Return to list of GRAIN functions at top
of this file
Return
to Main Index for the CDP System.
Return
to 'Groucho' Time-Domain Index
infile – input soundfile to process
outfile – output soundfile
code – a string such as adb:c indicating how grains are to be reordered.The example means use grains 1 (a), 4 (d) and 2 (b) in sequence, then-blen – maximum time between grains (Range: 1 to duration of infile)
begin this grain-jumping pattern again,
but start at grain 3 (c).
-lgate – required signal level for grain to be seen (Range: 0 to 1; the Default is 0.3)gate may vary over time.-hminhole – minimum duration of holes between grains (Minimum allowed value is 0.032 – the Default)
-twinsize – gate level tracks the signal level, as found with a window size of winsize milliseconds (Range: 0.0 to duration of infile)0.0 turns off tracking-x – ignore the last grain in the source
This is so much fun. Comparable with DISTORT SHUFFLE, GRAIN REORDER provides another way for controlled fragmentation. The illustration code used in the Usage causes a fair amount of fragmentation. But a code such as abcdefg:b steps through 7 grains in their normal sequence, then goes back to the 2nd grain and proceeds to the 8th etc., moving gradually through the whole infile in this way. The code gfedcba:b does the same thing, but places the grains in reverse order.
So you can see that you can play with the shaping possibilities to your heart's content, possibly making subtle correlations between the shapes made by the grains and patterns made elsewhere in the music.
End of GRAIN REORDER
Return to list of GRAIN functions at top
of this file
Return
to Main Index for the CDP System.
Return
to 'Groucho' Time-Domain Index
1 Repitch each grain in turn, without repeating any grains; on reaching the end of the transposition list, cycle back to its start.
2 Play grain at each transposed pitch, before proceeding to the next grain.
infile – input soundfile to process
outfile – output soundfile
transpfile – a file listing transpositions given as positive or negative semitone shifts. The maximum transposition is 4 octaves up or down.
-blen – maximum time between grains (Range: 1 to duration of infile)
-lgate – required signal level for grain to be seen (Range: 0 to 1; the Default is 0.3)gate may vary over time.-hminhole – minimum duration of holes between grains (Minimum allowed value is 0.032 – the Default)
-twinsize – gate level tracks the signal level, as found with a window size of winsize milliseconds (Range: 0.0 to duration of infile)0.0 turns off tracking-x – ignore the last grain in the source
This function is like GRAIN REMOTIF but only deals with transposition of the grains. Again, the transpositions are handled in terms of positive or negative numbers of semitones, so it is very simple to create the required transpfile. The transposition pattern / contour in the transpfile is transferred to the grains, such as narrow and wavy, or wide and jagged. This pattern can also serve as an important gestural or formal unifier in the composition.
The textfile can be written on a single line, with the values separated by spaces. Newlines are ignored. For example:
0 1 2 3 2 1 0 -1 -2 -3 -2 -1 0There may be some slight change in duration, proceeding from the source to the outfile, due to the effect of transposition on the final grain. Otherwise the grain onset timings should remain exactly as in the source sound.
The role of the two Modes is the same as in GRAIN REMOTIF, either taking each grain in turn, or applying all the transpositions in the textfile to each grain, repeating each grain as many times as there are transpositions.
Be careful with Mode 2, because an overly long textfile will create a very long outfile (and be very repetitive). However, the pattern is very clear, and a part of it may be especially interesting and able to be CUT and used elsewhere.
An ideal tool for creating staccato pitch figurations with clear contour shapes or even gestural potential.
End of GRAIN REPITCH
Return to list of GRAIN functions at top
of this file
Return
to Main Index for the CDP System.
Return
to 'Groucho' Time-Domain Index
infile – input grainy soundfile
outfile – output soundfile with grain timing re-patterned
timefile – must contain a list of grain-onset times in seconds. If any inter-grain time is reduced the minimum time allowed for a grain (0.032 sec), it will be reset to this minimum grain time.
offset – add this value in seconds to all grain timings – in effect, begin to apply the timefile pattern at this point in the infile.
-blen – maximum time between grains (Range: 1 to duration of infile)
-lgate – required signal level for grain to be seen (Range: 0 to 1; the Default is 0.3)gate may vary over time.-hminhole – minimum duration of holes between grains (Minimum allowed value is 0.032 – the Default)
-twinsize – gate level tracks the signal level, as found with a window size of winsize milliseconds (Range: 0.0 to duration of infile)0.0 turns off tracking-x – ignore the last grain in the source
This process re-times the start-times of the grains in the soundfile according to the pattern of times specified in timefile. The gaps between grain onsets can become larger or smaller in the timefile, but all onset times must increase – otherwise you would be trying to go backwards: e.g., 2.3 2.5 is OK, but 2.3 1.8 is not.
The offset parameter determines when after the start of the soundfile the timefile pattern is applied. If, for example, the offset is 2 (sec.) and there are grains before the 2 second point, you will hear these sound as normal, and then you will hear grains after the 2 second point patterned by the timefile
It is useful to run GRAIN FIND first, because it finds any grains in the soundfile and writes their start-times to a textfile. You can then edit this textfile to create a timefile with your own timing pattern.
GRAIN REPOSITION enables you to impose a rhythmic structure on the grains in a soundfile. Considerably more subtle effects are also possible. Trevor (Wishart) writes that it "provides a way to create a correlation between the fine attack structure and some other musical events. For example, the repositioning data can be taken from another grainy sound (using GRAIN FIND) or any other timing data, allowing the fine attack structure of a grainy sound to be synchronised or coordinated with another sound or sound sequence."
Here's a simple procedure:
- A timefile for the grains in a soundfile is quickly made with GRAIN FIND.
- Then GRAIN REPOSITION is run twice, e.g., with an offset of 0.25 the first time, and 0.33 the second time.
- The two files can be mixed with SUBMIX MERGE and
- You can then audition the result, hearing the double occurrence of the grains.
End of GRAIN REPOSITION
Return to list of GRAIN functions at top
of this file
Return
to Main Index for the CDP System.
Return
to 'Groucho' Time-Domain Index
1 Lengthen or shorten each grain in turn, without repeating any grains; on reaching the end of the time-multipliers list, cycle back to its start.
2 Play grain at each specified retiming, before proceeding to the next grain.
infile – input soundfile to process
outfile – output soundfile
multfile – a file of listing duration-multipliers to change the duration between one grain onset and the next grain onset.-blen – maximum time between grains (Range: 1 to duration of infile)
- the maximum value for a multiplier is 1000
- the minimum value for a multiplier is 0.001000
- If any inter-grain time is reduced the minimum time allowed for a grain (0.032 sec), it will be reset to this minimum grain time.
-lgate – required signal level for grain to be seen (Range: 0 to 1; the Default is 0.3)gate may vary over time.-hminhole – minimum duration of holes between grains (Minimum allowed value is 0.032 – the Default)
-twinsize – gate level tracks the signal level, as found with a window size of winsize milliseconds (Range: 0.0 to duration of infile)0.0 turns off tracking-x – ignore the last grain in the source
GRAIN RERHYTHM works in the same way as REMOTIF AND REPITCH, except applied to the timing of the grains only. Thus they retain their original pitch but begin at different times, according to the data in multfile. It may be that the grain lengths appear to change, because they begin to overlap one another: i.e., the distance between grain onsets changes, but the lengths of the grains are not meant to change.
The two Modes also work in the same way, either applying the data in the textfile to the grains sequentially, or repeating grains, applying all the data in the textfile to each grain in turn.
As pitch is not altered, the sense of movement is more directly perceived. The flow of the grains can be made to flex in supple or in strongly contrasting ways.
If there is a stuttering or rapidfire quality, it may be caused by the process not adequately separating the grains. If a group of grains are not seen as separate, they will behave as a single grain. It is essential with grain processes to ensure that all the grains are clearly separated (i.e., set the appropriate gate level). In some cases it will not be possible to get a totally satisfactory result (everything depends on the nature of the source). A little cosmetic editing may be required.
Also see GRAIN TIMEWARP.
End of GRAIN RERHYTHM
Return to list of GRAIN functions at top
of this file
Return
to Main Index for the CDP System.
Return
to 'Groucho' Time-Domain Index
infile – input soundfile to process
outfile – output soundfile
-blen – maximum time between grains (Range: 1 to duration of infile)
-lgate – required signal level for grain to be seen (Range: 0 to 1; the Default is 0.3)gate may vary over time.-hminhole – minimum duration of holes between grains (Minimum allowed value is 0.032 – the Default)
-twinsize – gate level tracks the signal level, as found with a window size of winsize milliseconds (Range: 0.0 to duration of infile)0.0 turns off tracking-x – ignore the last grain in the source
Unlike MODIFY RADICAL Mode 1 (which plays the source sound itself backwards, and thus can radically modify it), this process plays each original grain in the normal way (not reversed in time), but reversed in order. This is much like the idea of retrograding a melodic motif in traditional music (where the sounds themselves are not reversed).
The original source is thus always clearly recognisable in the output sound.
The forwards orientation of each grain means that the attack transient of the grains is retained. Even through the grains are generally very TW: Even though (Rather than 'even through') short, it is a tribute to the acuteness of our hearing that their forwards orientation makes such a difference in the recognisability of the original. GRAIN REVERSE, therefore, provides a wonderfully subtle variant of the original.
End of GRAIN REVERSE
Return to list of GRAIN functions at top
of this file
Return
to Main Index for the CDP System.
Return
to 'Groucho' Time-Domain Index
1 Mark where the iterative part of the sound is located.
2-3 The program attempts to find the iterative part, using envelope tracing.
infile – input soundfile containing iterative material
outfile – output soundfile with the iterations extended
stt – (Mode 1) - time of start of iterated material within source soundfile
end – (Mode 1) - time of end of iterated material within source soundfile
gate – (Mode 2) - minimum level in source soundfile for envelope tracing to kick in. (Range: 0 to 1)
usz – (Mode 2) - size of unit searched for, in milliseconds (15 for rolled 'rr')
ts – ('time-stretch') - multiplier that specifies how much to time-stretch the marked (or found) material
pr – guesstimate of the pitch-range of the iteration in the source, in octaves, or parts of octaves (try pr = 1). This is the pitch range of the (low-frequency) iteration, NOT the resonant frequency. E.g., for a rolled 'rrr', the frequency of the 'rrr' itself (even if unvoiced), and NOT the pitch of any sung note (if 'rrr' is voiced). (Range: 0 to 4)
rep – ('repeats') - the iterated material is extended by reusing individual segments in a randomised pattern. In this pattern, segment A may occur next to an identical copy of segment A, or not – see more detailed explanation below. (Range: 1 to 32767). Recommendation: try rep = 1 or rep = 2.
get – guesstimate of the number of iterations you expect to find (in the segment you've selected in the source – listen to the source). This helps to improve the accuracy of the search for iterated segment boundaries. In Mode 2, this is the minimum number of repeats you expect to find.
asc – random amplitude variation of the output segments. The value for asc multiplies the original segment amplitude by a random value in the range 1 to 1-asc. (Range of asc: 0 to 1)
psc – random pitch variation of the output segments. The value for psc transposes the pitch by a random value in the range -psc to +psc semitones. (Range of psc: 0 to 24)
skp – the number of found (iterative) units to skip before processing
at T – iterative ritard, after at seconds to event-separation T
by – ritard end point reached after a further by seconds; by = 0 gives no ritard
[-x] – (Mode 1) - keep rrr-extension only
[-s] – (Mode 2) - keep sound before the extended material
[-e] – (Mode 2) - keep sound after the extended material