On Monday, October 20, 2003, at 11:55 AM, Carl Edlund Anderson wrote:
> At 13:07 10-10-03, Michael Carlyle wrote:
>>> ps - speaking of custom kits, Mike C: do you have any notes on how
>>> you put together your ns_kit version? I figure the next best thing
>>> to a legally downloadable dbkit version of ns_kit would be a
>>> "recipe" DB users could follow so that everyone could get themselves
>>> more or less on the same page when swapping dbsong files that use
>>> such kits.
>>
>> This is a good idea, although the details should really be no secret.
>> I'll work on this later on when I get home and try to post something
>> to the list.
>
> I'm just going to ask about this again :) Mostly because I've messing
> around with kit-building and varying your version in different ways,
> but am interested in knowing how you went about getting there in the
> first place :)
Thanks for the reminder, Carl
Here's a synopsis:
After downloading and inspecting the zipped files from ns_kit, I
realized that the makings of a very nice Doggiebox project were there,
although the files seem directed more towards midi instrument users.
As I recall, I was immediately reminded about the true nature of
velocity variances. As you know, the volume an instrument produces is
only one of several complex variables that change with the velocity of
the strike. The ns_kit sounds contain many recorded samples for each
instrument, each one actually struck at a gradually reduced (or
increased, depending on your point of view) impact strength. The ride
cymbal, toms, and kick drums, in particular, sound almost completely
different than each's loudest velocity strike. Since I was aiming to
create lifelike fills and add subtle, human characteristics to my
patterns, it seemed like the perfect set of sounds.
Step by step, here's what I did:
Unpack and inspect the sounds. I used mostly Amadeus (an unregistered
shareware product I sometimes play with) and Bias Peak. to do my
overview and editing of the sounds. This was quite time consuming, but
I wanted to listen to everything.
Select the sounds I wanted to use. Rather than create a kit that
contained each and every sample provided by ns_kit, I thought I'd take
a representative selection that covered the entire velocity spectrum.
So, instead of having say, a ride bell that ranged from 1 through 8, I
chose, perhaps, 2,4,5, and 7 or whatever struck my ear as being the
most useful sounds. This took time. I didn't want to leave anything
important out, tonally, but I also did not want to overwhelm myself
with choices when using the kit. Also, as you may already have seen, a
DB kit becomes larger in direct proportion to the number and length of
the sampled sounds you feed it.
Edit the sounds for length and volume. After a quick test-run with a
new kit using some of the ns_kit sounds, I discovered that the volume
on playback was very weak as compared to a similar instrument from
Ben's Kit. I decided to pump it up a bit. I used Peak to open each
sound I wanted to amplify, and I determined a volume increase through
trial and error. After I figured out the volume I wanted to use, I
simply applied it to each successive sound file without even listening
to it as a time saving measure. It worked out OK, although there
remains a slight imbalance with certain instruments, to my ear. (This
is why I've repeatedly chimed in about the ability to export
multi-instrument files so I can mix them later).
I started to realize that some sounds had a very long decay that caused
the file size of the saved kit to grow rapidly. I realized that I was
reaching a point of diminishing returns. Would the very long, natural
decay of a ride or a crash or a snare spring rattle really add to my
sound's realism? I decided that much of this would be masked in a final
mix, so I began to shorten and fade to zero on a lot of the longer
sounds. I was able to radically reduce the file size of the final kit
in the long run. To this day, it remains annoying that I have to wait
for a ride cymbal to finish ringing on a pattern in order to get on
with programming.
Create kit graphics. Since my employment background includes the almost
daily use of applications such as Adobe's Illustrator and Photoshop,
this wasn't too difficult for me. I think I decided to stick with PDF
as the kit format. Organizing and deciding upon a naming convention for
the files is essential in order to locate these as you're ready to
apply one to a kit instrument. I used color to help me quickly identify
a sound in the pattern window. I also used some bits of text to help me
understand which velocity was used.
Put it all together. I'm sure you know the mechanics of building a kit.
It's pretty simple. I guess the key was my decision to create one
instrument per collection of velocities, rather than a whole bunch of
different instruments. It seemed a no-brainer to me. I chose the
loudest one for the default, and named the others in such a way that
their variance coincided with the velocity level of the ns_kit original
sample (numbers ranging from 1 to whatever). The only tricky part was
the hi-hat due to the inherent "muting" practice one must follow to get
sounds to shut off when others kick in. I do have a situation where the
ride cymbal gets muted by one of the hi-hats. I work around this for
now.
All in all, I'm fairly satisfied. I do adjust volumes and re-import
individual sounds back into the kit now and then. I also want to play
with the non-muted versions of the snare drum. He calls these rim-shot,
but I think this might be a misnomer.