KVR Audio

thagoocher · Post by **thagoocher** » Fri Oct 12, 2007 5:50 am

On some actual songs I've heard guitar samples that had tremelo on the sample itself, so that when you hear multiple notes, the volume peak on each note happens at a different time.

I'm not too bothered by these things, though. As I see it, one reason to use a sample over a real instrument (besides not having access to the instrument, of course) is to be able to do things with it that you couldn't do with the real thing. Authenticity is cool, I guess, but going out of your way to put back the string noise, fret buzzing, etc. (all things that musicians try to eliminate when playing a real instrument) is a little silly.

I guess it's like how video games go through the trouble of producing fake lens flares, an artifact that cinematographers generally avoid at all costs, or how animation will show characters in the background as being out of focus, until they speak, at which point the foreground goes out of focus and the background is in focus. This is to emulate a side effect of shooting in low light, where the opening in the lens must be opened up to get more light in the camera, producing a shallow depth of field. In both of these cases, real life is not being emulated, it's an emulation of life (film) that's being emulated. Weird.

Sam@Megablastic · Post by **Sam@Megablastic** » Fri Oct 12, 2007 8:58 am

After just finishing (the masters are now done at NI) a huge sample project of analog synths I suppose I should comment on this topic.

I don't like it when samples are normalized, it takes away some of the realism. When you have the lower volume sounds in a multi layer sample presets boosted to their maximum, all you do is boost the background noise with it or the room sound.

I made a point of having my library sound as close as I possibly could to the source material. Some presets have nearly 500 multisamples.

Another pet hate of mine in synth collections/samples is the lazyness of some developers i.e transposing a sample over 4 notes so it just doesn't sound realistic especially in the low notes. I've seen libraries where its like a sample for each octave. By the time you've transposed a sample down an octave you've lost half of the frequency response.

I also can't believe how bad and lazy some people are at looping samples.......oh and my biggest pet hate is:

Having 2-12ms (yes I've seen drum samples edited that bad)of space/silence before a sound actually starts. It's just down to lazy automated editing. Sampling is still an art form I believe, and not something you can leave to an automated bit of software.

koolkeys · Post by **koolkeys** » Fri Oct 12, 2007 9:09 am

Would you consider it lazy, or pretty standard, to not edit out the silence of a sample but rather set the start time right at the right spot inside the sampler?

For me, this would seem lazy because if people want to map themselves, they have to fix it. But then again, if you're just doing something quick and short on time, maybe it wouldn't matter.

Brent

pw · Post by pw » Fri Oct 12, 2007 9:13 am

electro wrote: A good library will not reduce or increase the volume of any sample.

Sam@Megablastic wrote: I don't like it when samples are normalized, it takes away some of the realism. When you have the lower volume sounds in a multi layer sample presets boosted to their maximum, all you do is boost the background noise with it or the room sound.

If I have 3 samples per key, assigned in velocity ranges 0-40; 41-90; 91-127. And, I have velocity->volume in the sampler at full range. And, I want to recreate some semblance of dynamic range from those three samples, the best way would be to match their volumes at the sample level.

Otherwise, the sensible use of velocity->volume would create an unrealistic dynamic range in the instrument as played in the sampler, when compared to the original recordings.

pw · Post by pw » Fri Oct 12, 2007 9:14 am

koolkeys wrote:Would you consider it lazy, or pretty standard, to not edit out the silence of a sample but rather set the start time right at the right spot inside the sampler?

For me, this would seem lazy because if people want to map themselves, they have to fix it. But then again, if you're just doing something quick and short on time, maybe it wouldn't matter.

Brent

I always meticulously hand edit start times of samples destructively, not in the sampler, for the exact reason you mention.

koolkeys · Post by **koolkeys** » Fri Oct 12, 2007 9:37 am

Well I asked because the libraries I'm working on and the mappings I've done, I've zoomed right in to the sample level in a wave editor and cut it right on the closest I could get. I think it's good practice, but I don't know if it's frowned upon to do otherwise.

Brent

koolkeys · Post by **koolkeys** » Sat Oct 13, 2007 2:25 am

Ok, I've got a question. Do you think it's really necessary to record both down pick and up picking samples in a guitar library? I mean, they sound the same, right? So are you just wasting samples by doing that?

Brent

Cordelia · Post by **Cordelia** » Sat Oct 13, 2007 2:35 am

koolkeys,
I would say that down strokes and up strokes on a guitar don't sound the same at all.
pw,
I think if you have only three velocity layers, you almost have to normalise and use volume scaling, but I would always prefer no normalising and more layers.

LlukaX · Post by **LlukaX** » Sat Oct 13, 2007 2:41 am

Sampling in general sounds weird to me :/ With proper use, physical modelling will get you better results, without the large memory footprint. for guitar try AAS string studio

Gregjazz · Post by **Gregjazz** » Sat Oct 13, 2007 2:45 am

koolkeys wrote:Ok, I've got a question. Do you think it's really necessary to record both down pick and up picking samples in a guitar library? I mean, they sound the same, right? So are you just wasting samples by doing that?

Generally the up pick has a little weaker sound. Of course, it depends on the player. More than anything, it's just getting rid of that terrible machine-gun effect.

Regarding the piano thing, it imitates the mic'ing of a piano (usually exaggerating it way too much, though). I think it's not as much as making the panning correlate to the position of the keys, but rather to give the piano a wider sound. At least this is how piano recording with mics started, but then I imagine it got corrupted/misinterpreted by keyboard makers who figured pianists like to hear high notes coming from the right channel and low ones from the left.

LlukaX · Post by **LlukaX** » Sat Oct 13, 2007 2:48 am

for piano, the best is pianoteq

you can even boost the key releases if you want to. also a lot less memory usage.

koolkeys · Post by **koolkeys** » Sat Oct 13, 2007 2:51 am

I'm all for giving the piano a wider sound and spreading the whole thing out. My problem was when you hit the top keys on the piano, it is panned hard right. When you hit the low notes, it's panned hard left. And everything in between is spread out. That doesn't sound like a piano, even when mic'd in stereo. At least none I've ever seen. But stereo width, yes. I think that is fine.

Now, as for up and down strokes, do you think that simple round robin samples will make up for the up and down strokes, and be a better choice anyways? Or do you really think it should have both? I'm asking you Greg because I trust your opinion on this.

Brent

Cordelia · Post by **Cordelia** » Sat Oct 13, 2007 2:52 am

Also, the upstroke, when I play guitar, has slightly more pick scrape to it as well, which I try to use to good effect.
Think about how straight eighth notes sound on bass using all downstrokes compared to down/up. Radically different.
Or how the intro to the Cars "Just What I Needed" would have sounded without those steady chunky downstrokes.

koolkeys · Post by **koolkeys** » Sat Oct 13, 2007 2:54 am

But does that matter in a sample library?

And also remember I'm talking about single notes here, not chords. With chords, you HAVE to have up and down samples to sound right, IMO.

Brent

Gregjazz · Post by **Gregjazz** » Sat Oct 13, 2007 3:05 am

Cordelia wrote:Also, the upstroke, when I play guitar, has slightly more pick scrape to it as well, which I try to use to good effect.
Think about how straight eighth notes sound on bass using all downstrokes compared to down/up. Radically different.
Or how the intro to the Cars "Just What I Needed" would have sounded without those steady chunky downstrokes.

Yeah, definitely pick scrape, too. Also when you palm mute, there's a lot of difference.

@Koolkeys: yeah, I'm annoyed with the hard panning, too. I like a bit of a wide sound, just so it's not mono-sounding, but the overly wide piano sound just does not sit well in the mix.

Regarding the upstroke/downstroke issue, I think you could get away with simple round-robin samples with cleaner guitar playing. I think palm mutes are the main thing. From what I've observed, the string tends to get muted more when you're executing an upstroke as opposed to a downstroke.

In any case, when sequencing and multiple notes are played at the same time, stagger them to represent an upstroke or downstroke motion. Another trick is if you have samples of a muted string being picked upstroke and downstroke, mix those in with the notes to give the notes variance. After all, it's mostly the attack sound of the sample that's the most important difference over subtle tonal differences.

Popular sampling conventions that don't sound realistic?