granular synthesis methods for ChucK

[Kassen]

Yeah, and by understanding it better it will help improve your music too!

[Junk Rhythm]

I know this is an old thread but here is a bit more to add to the topic. The following link goes to some granular ChucK code being interfaced by a monome.

http://post.monome.org/comments.php?DiscussionID=1011&page=1#Item_26

Might be of use for some down the road.

[Kassen]

No problem. I'd rather use old topics then re-do discussions all the time.

Looks&sounds cool! Did you consider using LiSa (see /examples/special/)? that would get you a lot more grains at less eford.

[Inventor]

Wow, I watched the video. You know, I have a matrix like that Monome in my Boolean sequencing program. I guess computer interfaces sometimes end up using a matrix of some sort. Thanks for the show!

[Junk Rhythm]

Actually, I didn't write the code in that link. I posted it here for others to look into for ideas.

It was because of that code that I finally dug into ChucK and tried to understand how things work. While digging around I came across this thread and also then discovered LiSa. I spent most of last evening messing around with LiSa which was a lot of fun so I will eventually implement something to that uses LiSa and the monome. I will definitely be back with some questions.

[Kassen]

Lovely, that's the best way to learn. Questions are welcome

[nununo]

Hi,

From what I understand LiSa is targeted at dealing with live recording. I want to do granular synthesis on an existing WAV. Can LiSa still be used? I don't seem to understand how.

Thanks,
Nuno

[Inventor]

SndBuff is a module that plays wav files. I haven't worked with SndBuff yet, but it seems to have position and rate controls. Here is the link to the SndBuff documentation:

http://chuck.cs.princeton.edu/doc/program/ugen_full.html#sndbuf

I suppose you could read a wav file in with SndBuff or the other way to read wav files which is WvIn, and route it to LiSa. Here is the link to the WvIn documentation:

http://chuck.cs.princeton.edu/doc/program/ugen_full.html#WvIn

[nununo]

Hi,

Can you please help me understand what do you mean by "route to LiSa"? Are you talking about copying the samples one by one from SndBuff to LiSa's internal buffer? Or are there ways to move the whole thing?

I'm sorry if this is a stupid question... but ChucK is totally new to me....

Thanks!
Nuno

[Inventor]

There are no stupid questions, only stupid people! Just kidding, of course. By "route to LiSa", I mean you could ChucK them together like this:

SndBuff sndbuff => LiSa lisa => dac;

Then when you read a wav file by using SndBuff method calls, you can simultaneously make LiSa method calls to record the wav file. After that you can play around with the sound that's buffered in LiSa.

[nununo]

I get it now. I'll give it a try. Looks like fun.

Although this seems to have the nuisance of, before being able to start playing with granular synthesis, having to sit through the whole WAV once while it is being ChucKed from SndBuff to LiSa, right?

In my case the source wave file is quite quite long: maybe over 10 minutes

[Inventor]

An easier way to read in the wav file would be to use WvIn, which has a simple interface. You could use the rate features of all three modules to read the wav file in faster, though that would decrease sound quality I suppose. Also, instead of reading into LiSa, you could use the playback features of SndBuff which include position, rate, and looping. That way you wouldn't have to wait for it to play the file as it reads it in to ChucK.

[nununo]

You mean something link this, right?

http://cm-wiki.stanford.edu/wiki/MultiGrain_Granular_Synthesis_in_Chuck

Ok, I think I will use this example as a start and tweak it into what I need.

Thanks again.

[kijjaz]

I've got an idea i'd like to try soon,
but if you guys want to try first, it'd be great.

I see that there can be some easy way to do granular synthesis with a simple SndBuf, an array of delays and envelopes.

If we connect one SndBuf to all the envelopes and the envelopes connect to the delays,
we surely can perform quite a lot of granular things.
although we don't have all the freedom in playing the sample,
but we can place many grains around quite easily.

I'll work with this idea if i have time.
It sounds quite simple in my head now, and I think we can apply this idea into many technics.
I hope we'll be testing it soon.

[Inventor]

What is the idea with grains anyway? Is it just playing back little snippets of sound from different parts of a wav file, all simultaneously? What little I've heard of grains, it just sounds like noise to me. What am I missing?

[Kassen]

[kijjaz]

Grains can be created from other kinds of sources other than playing directly from a crude sound sample,
One poppular way is to synthesize each grain. (not playing from a sampled sound)

For example, if the wave form in the grain is created from sine wave oscillator,
and the SinOsc's frequency increases continuously when the grain is played,
pass it through some kind of envelope to shape the grain's amplitude into a small piece of sound (maybe with fade-in and fade-out) ...

you'll get a small sinewave chirp by producing that kind of grains.
and multiple grains like that are lots of chirps surely.

It's a lot of fun creating synthesized grains because we can imagine up very different kinds of soundscape.

but it's not that better than making grains from sampled sound,
it's just another way to make use of granular synthesis technic ^_^
- - -

I've got a new idea!
I'm gonna make a class called "Sound Canvas" oh no! that sounds like Roland -_-"
Okay.. I'll think up a nice name hmmm....
DigitalTape! ..

okay, here's the idea:
this tape is a float array.
it's gonna hold the waveform, but functions can do many things to each part of the tape:
read (play sound), write/add, erase.

I still decide not to apply writing rate to the tape, coz I need much more time to study how to do effective interpolation.
(but i'll come up with some new ideas for that)

but with this, we'll have .. a canvas for painting sounds.
and so we can put some grain sounds to be played ahead in the future.
I guess this idea would become handy in this granular synthesis situation.

[Frostburn]

A nice idea kijjaz!
If you need any help with the math I can assist. I'd go by cubic interpolation with this one because sinc is more suited for steadily playing sounds.

[Inventor]

With your DigitalTape, you could set up rules for the grains. For example, as a grain plays the sound, it "eats" it, reducing the amplitude of that section of the tape. When a grain gets hungry it randomly jumps to another section of the tape to eat and play some more. Then you can have sun and rain which slowly grow the tape's amplitude. If a grain meets another grain, some interaction occurs such as they have lunch and eat the tape down to zero in that spot. If the grains run out of food for too long, they go dormant until more food arrives. Just a thought, thinking krazy like kijjaz!

[kijjaz]

Inventor: That'd be a fun effect to do! hahah.

Frostburn: I have this idea for doing the interpolation:
The "read head" and "write head" can shrink and expand in size, according to the read/write speed hahhahah

oh i have to go. i'll be back to continue on this soon.

[Inventor]

How about this? We take a sentence, such as "Port MAUI now please.", and chop it into individual word segments, or segments approximately the duration of one word. We then play back these grains in a clouded storm that begins with many disorganized voices of it and then over time they go into place gradually and form a repetition of the sentence.

So at first you hear a bunch of noise and then gradually over time the words form and finally the sentence. Might be a nice effect, in fact we could do it both forward and reverse, starting with a sentence and then breaking it into grains and then forming the sentence again.

Frostburn, what math would you use for that one?

[Frostburn]

[kijjaz]

Frostburn: Let's jam in this part.

I was wanting to make a tool that can read and write with varying speed also.
I once thought of a real tape:

if we speed up the tape, the energy transfered from the write head to the tape will be less per magnetic area.. but bigger in area
so i was thinking about resizing write area according to the speed.

then I think about writing as adding a signal x a distribution histogram to the cells of the float array..

aahhhhhhh I can't explain that well enough in English hahah...
ok..

-_-"

that eliminates the need to write more than once a samp..
but i then just had a new idea a few moments ago...
- - -

I think the above may not be that good for sound quality of the tape result after being written.. (but that still depends on the algorithm and how we calculate / choose the distribution type to "spray" value to the tape)

so I think about moving the tape head but not every samp=>now;
instead I think of samp/speed => now;

so now a good interpolation is gonna be essential.
also the interpolation algorithm would effect aliasing.
- - -

hmmm.. i can't explain well -_-
but i guess you've got some idea, and actually you must know how to do it efficiently. I'll see what I can do for now heheh.

[Frostburn]

I was just thinking about classes that make writing to and reading from float arrays easier when dealing with varying rates/speeds.

Attached is the code and a sound file it produced when I loaded a sample of yours truly saying "Please implement garbage collection."

---

digitaltape.ck
Description:	DigitalTape beta (only does linear interpolation now)	Download
Filename:	digitaltape.ck
Filesize:	7.17 KB
Downloaded:	293 Time(s)

---

---

garbage.mp3
Description:	An output of the program	Download
Filename:	garbage.mp3
Filesize:	140.63 KB
Downloaded:	752 Time(s)

---

[kijjaz]

Arrrrrggh.. totally cool. I'll come back to play with it soon! ..
Thanks. This is amazing.