Tips on making a multisampled instrument from my double bass

If you are new here check this forum first, your question may have been answered.
RELATED
PRODUCTS

Post

So, I have this big thing:

Image

I want to turn it into a couple of multisampled instruments. Arco and pizz with a few velocity layers for each, plus some miscellaneous scratchy, squeaky and percussive noises. Maximum realism isn't really my goal, just something that's gonna be quicker and more convenient and good enough for those times when I'm gonna be running the sound through a bunch of effects, or I just want to quickly lay down a scratch bass part that I can properly record later. I hope it turns out decent enough to be worth sharing and uploading somewhere, and if it does I can also do a couple more instruments.

I can get a reasonably clear sound recorded and make an SFZ file, but a few newbie questions before I start.


1. How wide a pitch range should I cover with each sample? I could record a note every half step separately, but would that be overkill?

2. Are extra round-robin samples worth it?

3. How many velocity layers?


So, basically, they all come down to the same thing: what's the optimum balance between the number of files that go into this thing and my own laziness?

Post

1. Semitones, of course, full range.

2. Absolutely, do at least 4 RRs, you don't want the machine gun effect, especially not with a pizzed double bass. For the arco you should at least do up and down bowing alternating.

3. 3-4

You'll have to edit hundreds of samples, but that's how it works, I'm just doing this for a 3,5 octave celtic harp and I have to kick myself every day to edit at least 80 samples which can take up to 8-10 hours. Recording is easy but cutting. levelling and filtering the samples is the hard part.

Post

As someone who's spent what feels like years working with sample mappings, I'll just say do not normalise the samples...

Post

So, about 1000 samples in total is needed to make something that's worth sharing for free? Damn, that is a lot of work. Still, it might be worth that kind of effort and I can break it down into smaller blocks. Maybe start with one velocity layer for arco. And make a guy who wants the squeaky bow noises do some of the work. And definitely no normalizing each sample individually, LOL.

Post

DSmolken wrote:So, about 1000 samples in total is needed to make something that's worth sharing for free? Damn, that is a lot of work. Still, it might be worth that kind of effort and I can break it down into smaller blocks. Maybe start with one velocity layer for arco. And make a guy who wants the squeaky bow noises do some of the work. And definitely no normalizing each sample individually, LOL.
You can start experimenting around and let your own ears decide what sounds good, sample every minor third, only do 2x round robins and only 2 vel layers maybe, then you don't feel so overwhelmed by the task :)
And as you want to share it for free you don't have to match any "industry standards" anyway.

Post

Yeah, if you're not going for out-and-out realism, even having a set of samples across the range in a couple of styles like you're suggesting would give people a base for their own creations. Even one sample of an instrument is useful to someone, after all, if it's just what they needed.

Post

I mean, yeah, if I wanted maximum realism I can just actually play a part on this bass. I want convenience for myself, and I kind of want it to sound like a more consistent, idealized version of my own playing. So I'll start smaller - less RRs and sampling every third sounds good, especially since I tune my bass in fifths so then every other sample from C1 to A2 is gonna be an open string. I'll probably record at least 3 velocity layers so the top one is nasty/growly/buzzy, though.

That would still be under 100 samples for arco and 100 for pizz, plus some misc noises - very manageable. And then I can build up from there, see how big I feel like making this thing. It might end up 1000 samples in a couple months, who knows.

Post

DSmolken wrote:I mean, yeah, if I wanted maximum realism I can just actually play a part on this bass. I want convenience for myself, and I kind of want it to sound like a more consistent, idealized version of my own playing. So I'll start smaller - less RRs and sampling every third sounds good, especially since I tune my bass in fifths so then every other sample from C1 to A2 is gonna be an open string. I'll probably record at least 3 velocity layers so the top one is nasty/growly/buzzy, though.

That would still be under 100 samples for arco and 100 for pizz, plus some misc noises - very manageable. And then I can build up from there, see how big I feel like making this thing. It might end up 1000 samples in a couple months, who knows.
I'd say you definitely do want at least 3 velocity layers. You could start with only 2 RR's, although what I would do is record say 5 or 6. This takes only a little bit more time than recording only 2. The more time consuming part is editing etc, which do only do for the first two, but if you decide to go back and add more, you've got the raw material.
(More than 4 RR's is probably unnecessary, but if you record 5 or 6 you can throw out the flubs).

Contrary to what an earlier poster said, I usually *do* normalize my samples. If you've only got about 3 velocity layers, you probably want some velocity->volume mapping, and if you've got that going on its a lot easier to get a consistent sort of velocity->volume curve over the whole range. For ultimate realism, of course, you would not normalize the samples, but then you'd a) need many velocity layers and b) probably have to adjust the velocity split points for each sample individually.

The other thing you probably want is note-off samples.

Good luck!

Post

one expects pizz will be more popularly sought after than arco. if you're looking for more to do, adding a bridge impulse (or sets with muted strings) would have application for a very few end users. hehe and string impulses :) ain't worth doing unless you do it the way you'll enjoy doing it.
you come and go, you come and go. amitabha neither a follower nor a leader be tagore "where roads are made i lose my way" where there is certainty, consideration is absent.

Post

mfb wrote:Contrary to what an earlier poster said, I usually *do* normalize my samples. If you've only got about 3 velocity layers, you probably want some velocity->volume mapping, and if you've got that going on its a lot easier to get a consistent sort of velocity->volume curve over the whole range
I'm usually dealing with thousands of samples and, with SFZ, it's trivial to get smooth mapping without normalising. Either way, you've still got to make a call on where the velocity->volume breaks occur, of course -- I just like the samples to give me some hint as to their relative loudness (particularly as I'm usually mapping someone else's).

Edit: Of course, with a drum kit, things are very different from an instrument with a pretty similar tone across all its samples - and I'm talking about drum kits.
Last edited by pljones on Tue Oct 01, 2013 7:29 am, edited 1 time in total.

Post

Thanks for all the extra tips. I'm gonna do this and take it fairly seriously. Got my bow rehaired, so it's a good time to borrow a good mic and do some recording.

I'm basically gonna record everything then see how many samples I really feel like editing (or can talk somebody else into editing). For my own use I want arco more than pizz, and a guy from a band I used to play in wants scratchy noises and muted strings and thumping the bridge and all that, since I've done a lot of that on stage. You're allowed to laugh:


Post

iirc years ago there was a cellist named fitch who released a big-at-that-time pack of wolftone kinda stuff. it was the kinda thing that ppl who post about samplepacks really seemed to like a lot.

i dunno if it's still around, so it may be ~redundant to do that for public benefit. or not :)
you come and go, you come and go. amitabha neither a follower nor a leader be tagore "where roads are made i lose my way" where there is certainty, consideration is absent.

Post

Your mastery of the Double Windmill technique is nothing short of amazing. Don Quixote would be proud! :D

I play double bass too, I have a couple nerdy questions:
- Are you doing mic, pickup, or a combo of the 2? I'm often surprised by how good a well-preamped pickup sounds for growly pizzicato.
- Fifths tuning...is that from a cello background, Red Mitchell, or something else? To me this makes sense for smaller instruments but seems like there's a lot of stretching/shifting when playing lines on one string.

Best of luck, interested in hearing it!

Post

I think Don Quixote would be really really angry, not proud, but...

I do have a homemade piezo pickup, but I think in all those videos I'm using a mic in the bigger venues, and just playing loudly unamplified in the smaller clubs like the double team/dry hump bits and the Edith Piaf songs.

Fifths tuning is from a mandolin family background, I played everything from mandolin to mandocello and even a cello banjo at one point. I got a doubleneck electric bass to tune one neck in fourths and one in fifths, then after a couple of years of that went "you know what... I barely use the fourths neck anyway".

Post

Got source audio recorded for the arco and scratchy bowing noises. 17 or so notes from C1 to A4, just up and down bows for round robins, five velocity layers. Also did some scratchy noises at the end so it all adds up to a bit more than a half hour of audio.

So, it might not be great, but it's gonna happen. I'll do pizz sometime later, maybe in an hour or two. Any last-minute advice before I do that?

Post Reply

Return to “Getting Started (AKA What is the best...?)”