Where to "start" a sample.

[Erick - BVA]

This has been one of those questions I've pondered quite a bit. It is especially relevant for guitar, piano and other libraries featuring a lot of attack in the sample start.
Do you programming the instrument to have the samples start as soon as a little bit of noise is present? In this case, you may get some piano key press sound being the very first sound you hear (as there is a tiny bit of time between the time you press the key and the sounding of the strings...however small of a time it may be). When using a special material to pluck a guitar, you may get the sound of that material brushing against the string before you get the sound of the guitar.
So do you edit the sample to start at the beginning of any sound? or do you edit it to start at the beginning of the actual sound of the instrument itself?
The reason I ask is that there is already a lag time with midi keyboards, so programming a sample to start at the slightest start of a sound will potentially lead to even more latency --essentially compounding the latency issue.
Is there a hard fast rule with this, or is it just another "it depends" scenerio?
Any advice would be appreciated.
My personal feeling is that you need to try to keep the attack, but to get it to a point where it is as close as possible to the beginning of that attack --if you cut the sample too early, then you have even more latency. You don't want any silence in the beginning. But cutting out the attack will lead to a lack of realism. But....I don't know.

Click to read more...

 

[Levitanus]

There're some users, who "definitellyheardelaofthreems". So, if You wont tell them that there's a gap before the attack and incrypt the instrument from opening - there is no question, attack has to be started from the beginning.

And, of course, the Gap has to be unisized)

 

[clisma]

Is it feasible to include a “sample offset” knob on the UI, linked to all groups and with a finite range?

 

[Vardaro]

Keep the pre-note noises.
They make all the difference between RealGuitar (v. good) and Ilya Efimov (outstanding).

I once cut them off in splicing a (live) piano recording: it then sounded like an electric piano.

 

[d.healey]

Keep the extra noise and provide a control to offset the position to taste. Another option is to link the offset to velocity so playing a higher velocity increases the offset.

 

[Levitanus]

provide a control to offset the position to taste

include a “sample offset” knob

I like to have "place for games" in stuff. But many controls for such things often brings to a situation when user-tweaks get much worse than built-in. The better to tweak everything "in the box". And, maybe, provide control in the very hidden place, where nobody finds it

 

[Old Timer]

I've just spent the afternoon sampling a (very bad) piano. I've been cutting up the samples and came up against just this issue. I was using Logic's snap to transient function (or whatever it's called). When I cut at the transient it didn't actually sound like my (very bad) piano any more. I found I needed to pull back the region to just before the main attack and I also found that each note was slightly different in how far I needed to pull it back.

So I think I'll have to adjust the sample start time for every single sample by ear once I put the samples into Kontakt (which will be a pain) but for me I think it will be the best way.

 

[Tod]

I've done a lot of sample editing and it's been my experience that every project will be different, at least in some way. It all depends on the instrument being sampled, the way it's miced, the number of mics, the way it's plucked if it's a plucked instrument, the room it's miced in if it's a miced instrument. Also noise can be a big factor.

As far as attacks go, they will vary, even from sample to sample. For myself there are 2 things to consider, making it too short or making it too long. If you cut it too short, you will get something like what "Old Timer" mentioned, basically cutting into the attack transient so that it's not natural. Cut it too long, then the sample won't play properly, worst case scenario, delayed, so that when played back on a keyboard it doesn't feel natural.

When you play a real piano, it reacts almost instantly, and the same has to happen with the piano samples, taking into account the latency of course, and at low latency, a sampled piano should feel close enough and natural, so that you can still play with feeling and expression.

When I'm editing the fronts of samples, I have a lot of tools (custom actions & macros) to make it quicker and easier, but I still edit every sample visually, to make that perfect cut.

 

[d.healey]

If you're editing in Reaper switch to the spectrogram view to make super accurate cuts. Very useful for cutting release samples too.

 

[Erick - BVA]

Wow, thanks for the great advice everyone!

 

[Erick - BVA]

If you're editing in Reaper switch to the spectrogram view to make super accurate cuts. Very useful for cutting release samples too.

Funny you should mention that. I discovered the spectrogram setting about month ago or so, and have not gone back since. It is much better for sound analysis than the usual waveform view.

 

[j.r.o]

I am currently working on a piano sample library and have been thinking about this...

If you're going for realism, shouldn't the attack time change based on velocity? If you hit a piano key softly, the hammer will take longer to hit the string than if you hit it hard, since mechanically speaking it will be slower.

Any thoughts or experience on this?

 

[d.healey]

I am currently working on a piano sample library and have been thinking about this...

If you're going for realism, shouldn't the attack time change based on velocity? If you hit a piano key softly, the hammer will take longer to hit the string than if you hit it hard, since mechanically speaking it will be slower.

Any thoughts or experience on this?

If you're recording multiple velocity layers then this attack based on velocity will already be in the samples.

 

[Erick - BVA]

I am currently working on a piano sample library and have been thinking about this...

If you're going for realism, shouldn't the attack time change based on velocity? If you hit a piano key softly, the hammer will take longer to hit the string than if you hit it hard, since mechanically speaking it will be slower.

Any thoughts or experience on this?

I think it won't change anything as far as sampling and editing goes. The sped up attack will be present in the recorded sound itself so no need to offset it in editing....I guess unless you want to exaggerate the effect. But I've been thinking about this since the last time I posted and it's clear to me that the initial attack may be one of the most defining characteristics of the sound. So getting all of the attack sound is crucial.

 

[Old Timer]

The initial attack is crucial to the character of a sound I think. But for me in a recent project there has been a trade off between keeping all of the initial attack and avoiding the sound being 'laggy'.

 

[j.r.o]

If you're recording multiple velocity layers then this attack based on velocity will already be in the samples.

Good point. I'm gonna go hang out in the think-before-you0open-your-mouth corner now

 

[Fredeke]

Because you can't live without hearing my 2 cents :

If you're making a multisample, it would at least be a good idea to be consistent across all the samples.

In general, keeping the whole attack will sound more natural. But then in extreme situations, a negative MIDI offset may be necessary in order to prevent perceptible (if not audible) delays.

(Ok, these are 4 cents.)

 

[Erick - BVA]

Because you can't live without hearing my 2 cents :

If you're making a multisample, it would at least be a good idea to be consistent across all the samples.

In general, keeping the whole attack will sound more natural. But then in extreme situations, a negative MIDI offset may be necessary in order to prevent perceptible (if not audible) delays.

(Ok, these are 4 cents.)

Yes, I've had this issue recently with a library. The amount of time from pressing the key to hearing a note needs to be consistent from note to note, otherwise you lose the feeling of realism because of the varied delay from note to note. And it's just extremely annoying to play, and hard to keep it in tempo. It needs to respond consistently from note to note...at least to a point where it's not perceptible (as you said).

 

[Fredeke]

As for the sound of a finger hitting the piano key : it occurs a tiny bit before the note. Just as your finger hits the MIDI keyboard a tiny bit before the MIDI message is registered. So, for more realistic timing, I would cut out that sound. But for more realistic sound, I may want to keep it. To get the best of both worlds, I would keep it, but then I would very slightly move the MIDI track backwards in time after recording. That is the ultimate solution if you really like splitting hair