Welcome back to Intense Tech with Defense Mech! In this lesson we’re going to cover how to use DOTCNT‘s LSDj Wave Cruncher to take samples and generate your own custom wave synth instruments that you can patch into LSDj save files and songs!
You may remember the previous Intense Tech lesson featuring the LSDJ wavetable import tool that 4ntler created, which allows for patching wavetables directly into an LSDj save file or lsdsng file, such as those from the LSDj wave synth instrument library. But how do you generate a wavetable synth from a sample? Today we’ll answer that question as we deep dive into the incredible work of DOTCNT, the LSDj wave cruncher!
I had the pleasure of meeting Ada, aka DOTCNT in France last year, and I immediately praised her for the incredible work she had done to provide the LSDj community with this tool. In essence, the wave cruncher is a way to automate the downsampling of a single sample into the format that works with the LSDj wave synth, essentially turning it into a wavetable sampler. This allows you to play notes, bend the pitches, arpeggiate chords, and so on, in ways that you can’t when playing samples from kits in the LSDj ROM. This lesson will be incorporating a lot of previous knowledge from past Intense Tech articles, so feel free to refer back to those if you need to!
The original tool requires you to install node, Python 2, and several libraries for each, but I’ve also compiled binaries that should run in the command-line/terminal environment on Mac, Windows, or Linux machines without involving any setup. (If you need a refresher on the command-line/terminal, feel free to check the article I wrote about liblsdj as well.)
There are a few more important requirements you will need before you can start crunching some waves. First, you will need your instrument sample. Ideally this should be a .wav file that contains just 1 note. It’s also important that the note be tonal – that is, it needs to be a constant pitch. Think of a musical instrument that plays one note, and you’ll have the idea of the kinds of sounds that will work well with wave cruncher. Sounds that work especially well are those with lots of overtones like bells, FM e-pianos, steel drums, and so on. Sounds that won’t work well are sounds that are especially noisy, or non-tonal such as snare drums, train whistles, vibra-slaps, or samples from songs by the Red Hot Chili Peppers.
For this example, I’ll use an e-piano sample.
FM E-Piano sample
Next, you will need the wave cruncher binary. Download the appropriate version for your platform and place it in the same folder as any samples you want to crunch, and open a command line terminal to that folder.
When you run wave cruncher, you’ll provide it with certain parameters or arguments, like so:
crunch [SAMPLE.WAV] [NOTE|FREQUENCY|auto] --flag
“Crunch” is whichever binary you would call on your system, for Windows we would call “crunch-win.exe” followed by the filename of the instrument sample we want to use such as “epiano.wav”. Following the sample is either the note such as “A4”, the frequency in Hz such as “440”, or simply the word “auto” which will attempt to auto-detect the frequency of the sample. Auto-detection is very nice when it works, but it may sometimes fail to detect the pitch of some overly complex waveforms, and it can take around a minute or so to work (or fail), which can be time-consuming.
You can feel free to run the cruncher at this point, and if all goes well, you’ll be rewarded with the output of a .snt file which can then be patched into a save file using the liblsdj wavetable patcher.
Let’s take a look at the output when we run
crunch-win.exe epiano.wav auto
$ crunch-win.exe epiano.wav auto
Frequency detected: 136.47859664570356 Hz
Saving data as epiano.snt... Done!
Successfully output epiano.snt!
At this point, we can patch “epiano.snt” to our LSDJ save file. However, we can also provide additional flags to finetune the wavetables.
The first optional flag affects which wave cycles the cruncher puts into the wavetable
output. By default, it takes the first 16 wave cycles. But this may not really capture the
entirety of the sound of the sample, especially if the sample is long. So the first flag,
--linear“, takes a linear sampling of wave cycles from the
entire waveform from beginning to end. You can think of this as slicing the entire sample
evenly into 16 equal slices, and taking 1 wave cycle from each slice.
The second flag, “
--exp“, mimics a kind of exponential sampling. In other words,
it takes more wave frames from the beginning and fewer wave frames from the end. This can be
useful if you want to capture more of the nuance of the attack of the sample and less of the
decay or release. It’s probably worth playing around with these options as they can
drastically change the sound of the resulting wave instrument.
The next optional flag. “
--normalize“, affects the volume of the
sample. Adding this flag will maximize the volume of the sample before it is crunched. This
is often advised if the sample hasn’t been amplified because normalization generally
results in better-sounding synths. Your samples may already be normalized, in which case
this flag is safe to omit.
The third optional flag is “
--channel=1” for stereo wave files. Since the LSDj wave synth is
monophonic, the wave cruncher has to select either the left or right channel of the sample
to crunch. By default it uses channel 0 (left), but if you prefer, you can set the flag to
channel 1 to crunch the right channel instead.
The fourth optional flag is “
--output=filename” which allows you
to specify the output of the .snt file. By default, wave cruncher uses the filename of the
wave file sample, but this can get confusing if you want to test results of different
parameters for the same sample. By specifying
--output=bell-exp.snt” for instance, you can specify that the
resulting .snt used the –exp flag, rather than the default or linear sampling flags.
In short, it is just designed as an option to help keep things organized.
Lastly, there is an optional “
--analyze” flag which is simply
designed to analyze frequency. It will not output a .snt file. Think of this as a test mode
– you can run the cruncher with –analyze to make sure that the pitch
auto-detection doesn’t fail. This can be helpful when batch processing samples just
to ensure that the process will run smoothly without hiccups – if pitch
auto-detection fails, the sample won’t be crunched. Auto-detection can also add about
30-60 extra seconds to process the wave file. Since the
--analyze flag outputs
the frequency that it detects, frequency can then be provided in subsequent crunches to
speed up the process.
Let’s run the cruncher a couple more times with our epiano sample and test the results. Since it took about 45 seconds to detect the frequency, I’ll provide the value instead of using ‘auto’ to speed things up, and see what sounds result when using the different flags.
$ crunch-win.exe epiano.wav 136.47859664570356 --normalize --linear
Sample rate: 44100
Taking frame every 60 cycles
Saving data as epiano-lin.snt...
Successfully output epiano-lin.snt!
$ crunch-win.exe epiano.wav 136.47859664570356 --normalize --exp
Sample rate: 44100
Saving data as epiano-exp.snt...
Successfully output epiano-exp.snt!
Once you run the cruncher, you can then follow the instructions here to patch your .snt files into your save files and lsdsngs! Keep in mind that you will have to set the Play, Length, and Speed parameters in the wave instrument appropriately. Depending on the tempo, try it with Play Once, Length 3, Speed 3, and adjust from there. If the result sounds good, please consider contributing them to the LSDj synth instrument library! Let’s check the results of our instruments below. I’ve patched each instrument into Instruments 1, 2, and 3, and Synths 1, 2, and 3 respectively. Chains and Phrases 01, 02, and 03 will play these respective sounds.
Each Wave instrument may need different parameters
I hope that this article gives you some insight into the wave cruncher and encourages you to try crunching some of your own wave samples!
Note: traducción al Español por Pixel Guy encontrado aquí.