Sunday, 14 October 2018

ProbablyR v011 - Updated MaxForLive Probabilistic Sequencer. for Ableton Live

I have just released a new update to my Probably MaxForLive generative, probabilistic sequencer - this is ProbablyR, and it adds a lot of new functionality, although not everything that was requested has yet been added - there are more updates coming.


The above graphic shows a summary of the main changes:

Memory slot grid


(This memory grid is in 'Setup' mode, only four slots have been filled, and there are no white spots because they don't do anything in 'Setup' mode! They just recall the contents of memory slots when in 'Play' mode.)

The memory slot grid now has 16 steps. Previously this only had 8 steps, and so whilst you could store 12 memory slot (the square grey boxes in the 'black' section), you could only control which one was playing using 8 steps. Using the maximum length of 16 steps, and if you select the maximum number of repeats (16), then you have 256 (16x16) bars before the sequence repeats. Having more steps than memory slots means that it is easier to use a couple of memory slots to allow organisation of slots and their contents.


(This memory grid is in 'Play' mode, and there are four filled memory slots (on the right hand side). The Length is set at four, and so steps 1 to 4 will repeat again and again until you stop Live. The four white spots select memory slot 1, then 2, then 3, then 4, and then 1 again. ProbablyR is currently playing step 2...)

The workflow is intended to consist of:
0. When developing a new sequence, use the 'Clear' button in the lower right hand corner of the 'Length' section (it is deliberately kept well away from everything else). This will wipe the contents of the memory slots forever, and they cannot be retrieved. Only do this if you want to start from a clean set of memory slots.
1. Developing memory slots using the 'Setup' mode, and using Shift-Click into slots to store a snapshot of the coloured sections for each major change.
2. Using the 'Play' mode and the memory slot grid to determine the order in which the slots are played back, with the shortest sequence being a length of 1 bar, repeated once, and then starting again (and the longest being 16 bars, repeated 16 times, and then starting again).
3. Jumping back to 'Setup' again to make an edit to a memory slot. (If you don't do this then the recall of the memory slots will keep overwriting your changes! Whenever you find that your grid edits are being overwritten, check that you are not in 'Play' mode!)
4. Using 'Play' mode to play the memory slot grid.

I'm working on ways to make this workflow more intuitive...


Here is a compressed screen shot of that 'memory clear' action (step 0 of the workflow above). Notice the round red button with an 'X' in the lower right hand corner of the Length section, and the red square memory slots on the right of the Memory section. In the memory section, red squares mean 'empty', grey squares mean that the slot has something in it, and a white square is the memory slot that is currently being played.

Remember that the memory slot grid has the same probabilistic control as all the other grids in ProbablyR, so if you have two white spots in a vertical, then they will be chosen on average 50% of the time. If you have three, then 33% each, and so on. You can use just single spots if you want to know exactly what will happen, but using more than one white spot gives you controlled randomness...


So in the above screenshot, the second set of 4 bars is playing the recalled memory slot 2. When the 9th bar starts, the two white spots in the vertical line (one to the right) will select either memory slot 1, or memory slot 3 - with a 50% probability. After another 4 bars, then the two spots in the right-most vertical mean that either the 3rd or the 4th memory slot will be played - again with 50% probability. So there are 4 ways to play the memory slots: 1213, 1214, 1233, or 1234. Because there are two sets of 50% choices, then each of these possible ways will happen about 25% of the time.

Remember that a memory slot contains all of the settings in the coloured sections (except for the step length and the memory reset!). So those 4 ways of playing are actually 16 bars where there could be additional probabilistic operations set...

Order grid

The 'Order' grid controls the order in which steps are played. The default grid is a diagonal line from the lower left corner to the upper right hand corner. You can set this by clicking on the '1' round button on the right of the purple 'Order' section. The steps will then move from left to right, and the sequence plays 'forwards'.

Forwards

The opposite to this is set by clicking on the '-1' round button, and this puts a diagonal line from near  upper left corner to near the lower right corner - the exact positioning depends on the length of the sequence. In the screenshot the length was 16 steps, and so the last two steps in the grid are not used. The sequence now plays backwards.


Backwards

The round button with '4' in it is a special case that sets just every 4th white spot, and is typically used when you are working with a sparse number of notes per bar. The screenshot shows this situation...


'4': every fourth note

There are 5 white spots because the full grid is actually 18 steps in length, and so an extra white spot is required. In the case of 16 or less steps, then the 5th spot is ignored.


Forwards and backwards in the same bar

By setting the order grid so that it goes forwards for part of the bar, and backwards for the rest (16 steps are shown here), then the order of playback of notes will go forwards for part of the bar, and backwards for the remainder of the bar. This type of control over the order of playing of steps is quite unusual in sequencers (software or hardware). 

Order randomisation

Instead of just forwards, backwards, and any combination of those, there is also an experimental randomisation function that allows the order of the steps to be randomised, every bar, every 2 bars, every 3 bars, etc. This only works on the 16-step grid at the moment (the code to produce different random grids is not straight-forward!), but it has two controls: the small square button containing 'N' or 'R', and the small number underneath it, which sets how many bars have to be counted before a new randomisation is generated. When the button shows 'N', then a new randomisation set is produced every so many bars, and when the 'R' is shown, then no new randomisation is generated (so you can leave one in place if you want). The large white number is the bar count.


A random order grid held by the 'R' button setting.


The random grid about to be replaced (the button has just been clicked to change the 'R' to an 'N' ('No randomisation')) - the bar setting is '1', and the bar count is '1', so at the end of this bar, a new random order grid will be produced.

Note repeat

The order grid can also be used as a shortcut way of repeating a note. If you set up the order and pitch grids like this:


...then the dark highlighting that I have added across the order grid is actually a time plot of the note in step 4 that is highlighted by a vertical bar that I have added in the pitch grid. So the second white spot causes the note to be played again. This can be a very quick way to add little extra 'busy' notes to a sequence, and can be combined with the probability controls via multiple vertical spots. There are lots of ways to exploit this to make interesting patterns with many variations.

Step length

Probably the most important change is the step length. Instead of being locked to 16 steps, you can now choose a step length from 9 steps to 18 steps. This equates to time signatures of 4.5/8 to 9/8. If you put a copy of ProbablyR in more than one track inside Live, then you can have different step lengths for each track. Combined with the probability settings , this can be used to create music with very large numbers of variations (and a long time before it repeats the same music).


Here is the length box, with the length control set to 16. The range of the control is from 9 to 18 steps.

Note that the length control is NOT stored in the memory slots. Experimentation with the beta versions showed that this was not a good idea because it caused nothing but confusion.

That completes this update report for ProbablyR. At some stage, I will gather all the material in these blogs posts on Probably and edit it into the user manual...

Tutorials and further information

The 'Probably' sequencer is a complex plug-in with an unusual user interface. The following online resources are recommended until I produce a full user manual:

http://synthesizerwriter.blogspot.co.uk/2017/07/probably-antidote-to-step-sequencers.html

http://synthesizerwriter.blogspot.co.uk/2017/09/probablyz-tutorial.html

https://soundcloud.com/martinruss/probablyz-tutorial-demo-01

http://blog.synthesizerwriter.com/2017/10/probablys-tutorial-using-newly-added.html

Getting ProbablyR_mr_v011

You can download ProbablyR_mr_0v11 for free from MaxForLive.com.

Here are the instructions for what to do with the .amxd file that you download from MaxforLive.com:

     https://synthesizerwriter.blogspot.co.uk/2017/12/where-do-i-put-downloaded-amxd.html

(In Live 10, you can also just double-click on the .amxd file, but this puts the device in the same folder as all of the factory devices...)

Modular Equivalents

In terms of modular equivalents, then reproducing this functionality in my modulars looked like it would require several sets of four 8-step sequencers (plus additional logic modules), or it would require the use of advanced sequencers that I can't count as 'basic' modules, so I would rate this version as being about 140 ME. This is not something that I would want to make for real in a modular.











Tuesday, 9 October 2018

Residuals Using Simple Separation - Seminar at Synthfest UK 2018

On Saturday the 6th of October, 2018, at 2:45, I presented a seminar about my method of using residual samples to do sound synthesis at the 2018 Synthfest UK event.


The title of the seminar was 'Sound Synthesis Using Residuals', and it described some research that I have been doing.



I promised to make the Nyquist filters for the Audacity audio editor available, and here is the link:

Filters for Audacity

The two filters look like this:


The 'Harmonic Extractor' filter outputs only the harmonics of the input, if the frequency of the narrow peaks matches the pitch of the input.

The 'Harmonic Rejector' outputs everything except the harmonics of the input (the 'inharmonics'), if the frequency of the deep notches matches the pitch of the input.

The shapes of these filters are not your ordinary synthesiser staples like low-pass filters. The shape of the harmonic extraction filter is sometimes called a 'teeth' or 'tooth' filter, whilst the harmonic rejection filter is often called a 'comb' filter.

Harmonics and frequencies are related in a not-quite-straightorward way. The pitch frequency is called the Fundamental, and is often represented by the symbol 'f' (for frequency). The first harmonic is at twice the frequency of the fundamental (2f), the second harmonic is at three times the fundamental (3f), and so on until your ears run out of range, or a digital representation of the audio gets close to half of the sampling rate and starts to be filtered out to avoid aliasing. Usually, all you need to know is that you add one to the harmonic to know what multiple of the fundamental frequency it is. So if the fundamental is 100Hz (f), then the 1st harmonic is at 200Hz (2f), the 2nd harmonic is at 300Hz (3f), and so on.  

What these filters do is to align the peaks or notches so that they match the harmonics of the input sound, and they then either extract or reject the harmonics in that sound. (This process is not perfect, of course...)

The RUSS synthesis technique uses these two filters to extract the harmonic part, and the inharmonic part, of the input sample. Once you have these two parts (called residuals), then you can use conventional S&S synthesis to create sounds using them as raw material. Separating the sample into two residuals lets you do all sorts of interesting edits...

More information on the RUSS synthesis technique can be found here.