9 ways to NRT, part 2: NRT and Csound

2014-12-20

Welcome to Part 2 of 9 Ways to NRT

Today, I'm going to show how NRT can be used to play a synthesizer in Csound.

Recently, I discovered that the '-Lstdin' flag lets Csound read from standard input. This means that NRT can pipe stuff to Csound without ever needing to write anything to file.

NRT can't do this directly; it needs an intermediate filter to convert NRT output to a Csound score. Luckilly, this couldn't be easier with a little help from our old pal Awk.

Introducing Awk

Awk is a scripting language designed to extract data from text files, one line at a time. From there, Awk can do stuff to those values. Awk can either read from a text file or from stuff piped from standard input; we will use the latter to directly send NRT's output to Awk.

NRT output is designed to be fed to Awk, so it is a good practice to use Awk as an encoder before piping NRT data to external programs.

Creating a Csound Score with Awk

Here is an example of NRT creating a Csound score with Awk:

$ echo "drm" | nrt | awk -F',' '{print "i1",$1,$2,$3 + 60}'

i1 0 1 60
i1 1 1 62
i1 2 1 64

We are piping NRT notation into nrt using echo. Then we are piping the output of nrt into the input of awk. Awk then prints the formatted Csound score, adding 60 (C4 in midi) to the note number, which puts everything in the key of C.

This is a valid Csound score, but this program will only write melodies at 60bpm in the key of C. We can modify our program to have variable tempo and key (note offset). The program is still technically a "one-liner", but it has been broken up into 2 lines:

$ echo "drm" | nrt | awk -F',' -vTEMPO=120 -vBASE=60 \
 'BEGIN{ts=60/TEMPO} {print "i1",$1*ts,$2*ts,$3 + BASE}'

i1 0 0.5 60
i1 0.5 0.5 62
i1 1 0.5 64

The Csound Orchestra

To match the generate score, the Csound orchestra needs an instrument with an ID of "1" that reads in a midi note number (the 4th p-field).

The instrument below in a simple karplus-strong instrument being sent to a reverb. You can copy and paste this into a file called "pluckrev.csd" or you can download the file here.

File pluckrev.csd:

<CsoundSynthesizer>
<CsInstruments>
sr	=	96000
ksmps	=	1
nchnls	=	2
0dbfs	=	1

alwayson "reverb"
gaRev init 0
instr 1	
icps = cpsmidinn(p4)
a1 pluck 0.1, icps, icps, 0, 1
a1 = linsegr(1, 1, 0)
outs a1, a1
gaRev += a1  0.3
endin
instr reverb
aL, aR reverbsc gaRev, gaRev, 0.97, 10000
clear gaRev
outs aL, aR
endin
</CsInstruments>
</CsoundSynthesizer>

Piping it all together

Everything can now be piped together in a single bash command. I've changed the melody here to make it more interesting:

$ echo "drm2tsd" | nrt | awk -F',' -vTEMPO=120 -vBASE=60 \
    'BEGIN{ts=60/TEMPO} {print "i1",$1*ts,$2*ts,$3 + BASE} ' |\
    Csound -Lstdin -odac pluckrev.csd

The output of Awk is being piped into the input of Csound. Depending on your operating system and driver setup, your Csound command line flags may vary. For instance, this is what I had to put in to get to get Csound to run on my Linux machine running Jack:

$ echo "drm2tsd" | nrt | awk -F',' -vTEMPO=120 -vBASE=60 \
    'BEGIN{ts=60/TEMPO} {print "i1",$1*ts,$2*ts,$3 + BASE} ' |\
    Csound -Lstdin -odac:system:playback_ -+rtaudio="jack" -b 2048 pluckrev.csd