Note-off velocity?

I think in the MIDI specification this wasn't mean so much with pianos in mind, if at all, but rather for synths to have just one more dimension for expression. I would wildly assume that this may come useful with ethnic types of sounds where the pitch may suddenly blip at the end of the note and depending on the velocity with which you release the keys.

Now, to pianos. If you release the keys slowly (and the damper pedal is not pressed at that moment), the dampers would gradually dampen the string which means the tail of the note would be longer and not abruptly cut-off. On real pianos that's sometimes called finger legato where you play the next note while you're still holding the previous key, or while you're releasing it slowly, so that the notes connect smoothly and are almost blended like with a pedal.

On my Yamaha N1X this effect is supported, although it's very subtle and I have to specifically listen to it to notice it. But I guess as with many other things in life and in audio, subtle effect are nevertheless cumulated in our perception and brain to result in a better ultimate satisfaction at the end.

BTW, the Yamaha implementation is not just about the velocity of the released keys. It actually detects the actual position of the keys, hence dampers, depending on the key depth. Or in other words it's a half-damper effect linked to keys too, on a per key basis, not a global one such as the half-pedal effect which affect all the dampers together. Yamaha implementation is one step further since the release velocity assumes a single event that ended: a release. Whereas the Yamaha implementation will support the key being released only slightly to a middle position after which it can be pressed again to restore the natural sustain of the string from that position onwards.

In reality on a well regulated grand, there's not such a discernible effect and the dampers are mostly instantaneous, but yet, the effect exists to a degree. Listen to real acoustic piano recordings and you will hear little buzzes from time to time: these are caused by very slow release of the keys.

The "note off" event tells that (in case of a keyboard instrument) the key was released. That's often enough.

The "velocity" then is the speed at which the key was released. In an acoustic piano it affects how the note dies out. Some digital pianos also use it to reproduce a similar effect. Yamaha calls it "Smooth Release".

But how exactly do they do it? And doesn't the data come only after the physical key release event has already happened? Maybe the two lowest sensors are used(?) Do some dual sensor keyboards also have it?

I have never tried it on a digital piano.

I think the sonic effect of a slowly dropping damper is usually simulated.

But don't some virtual instruments have note-off samples? Are those the sounds from slowly dropping dampers?

Release samples are not about dropping hammers, these are samples of the release portion of the piano sound - with the notes being dampened to silence on a piano. For that reason there are no release samples of the upmost octave.

Deleted dropping hammers

dropping DAMPERS, not hammers.

vagfilm I see it now. I probably thought of these useless hammer dropping samples built into some digital piano.

I think, that's it for me. Time to drop out off all this pointless tech talk and focus on the art of playing the piano itself. In the future I probably will make good use of the forum feature to hide threads about "note-off velocity", "half-pedaling" (which is not at all about the piano technique itself) and all the other stuff, which doesn't bring anyone closer to being a better artist.

Argh. After building a half damper Arduino to experience half pedaling, now I need a new controller that supports release velocity?

Seriously though, do sample libraries support this or is it mostly limited to internal DP sounds?

At least Pianoteq supports it (although by default it is set as a flat curve). Some VSTs (VILabs, I believe) claim to have release samples at different speeds, but I am not sure if they use the actual note-off velocity or simply the time difference between note-on and note-off (so, very short notes are assumed to be staccatisimo regardless of the velocity of the key on the way up…) I'm no expert but I think you can script this similarly to the half damp: it will affect how abrupt the dampening will be. AFAIK, Roland, Yamaha and most Kawai pianos send note-off messages with velocity.

Its been pretty fascinating, this process of discovery for me - translating what I have been doing for years subconsciously on my acoustic into MIDI continuous controllers. I guess I'm so used to the standard DP for so many years, I'd just accepted that a DP will never be an acoustic. I still believe that, but certainly DP's are better than I thought, and I need to experience it to really understand it.

I mean, for years as a kid I wondered why I "ride the pedal", and then I forgot about it. But now I know I'm just controlling the brilliance, sustain, and preventing "muddiness" unconsciously. Probably the same with release velocity - I probably utilize that too without knowing it. I'll have to try.

Thanks for all the contributions here. Going back to CG's point that note-off velocity is not a piano concept but a synth concept, this makes a lot of sense to me, the more so now after having read the posts above, and thought more about it.

The process of letting go of the key in an acoustic can never be captured meaningfully by a single number, like a velocity, because this process can be steered arbitrarily by the player. Example: the key is pressed to the bottom, and now you release it halfway up. In response the damper is lowered somewhat. Now you might just push the key back down to the bottom again, and the damper would be raised again. This all takes place within one keystroke, as long as escapement is not re-activated. In fact you could modulate the damper position by moving the key up and down as many times, and as fast, as you like. This insanely complex behaviour can never be captured in a single number (of course doing this sort of thing would also be a bit insane, but that is besides the point 😆).

QuasiUnaFantasia Example: the key is pressed to the bottom, and now you release it halfway up. In response the damper is lowered somewhat. Now you might just push the key back down to the bottom again, and the damper would be raised again.

The dampers (with either foot or finger) move by a relatively large amount, say about 1cm. The actual damping effect, buzzing and everything we are talking about takes place in a relatively small area: from when the felt start touching the string (when damping) to when it is completely pressed against it -- I think that's about 1mm. Plus, the first position depends on the volume of the note (louder notes are caused by strings vibrating a larger amount). Pretty complicated to keep all of this in a software. Plus, even in a acoustic grand, the dampers are seldom correctly regulated with the pedal: i.e. some notes are damped earlier and some are later. Try it:

• press the pedal and keep it
• press and release a note
• slowly release the pedal but not completely, only until it "just starts" dampening that note (at the volume at which is playing)
• keep the pedal still
• try the same note at the same volume again, and adjacent ones: likely they will behave differently (this is a regulation defect)
• keep the pedal still
• try one note at different volumes: the louder ones are damped, the softer ones are not (this is an acoustic characteristic which I doubt any software correctly reproduces -- but then I am not sure any artist take advantage of)

Bottom line: should we care about note_off velocity as way to get a digital piano become a better approximation of an acoustic piano? I dunno.

I'm perfectly happy about the note-off velocity always being zero. My problem was that I didn't understand the purpose for this part of MIDI. I suspect that various makers of VI's do use this velocity for something, but I don't believe they use it for something meaningful in the context of an acoustic piano (at least, that is how I view it now).

Some digital pianos don’t even send the note-off message and instead send a note-on with value 0 to denote that the key has been released.

CyberGene Some digital pianos don’t even send the note-off message and instead send a note-on with value 0 to denote that the key has been released.

In fact, note-off encoded as note-on with velocity zero is still the norm in the majority of midi files found in online repositories… But the current line up of mid and high level digital pianos, all send note off (or at least has that feature in their midi specifications), with a few surprising ommisions…

When I last tested with a MIDI app, on the NV-10's sound engine the note off velocity seems to affect the volume of the release sample as well as the volume of the hammer noise. I couldn't tell any difference in the nature of the sample itself.

I never tested my VSL D274 or even the garritan for note-off differences (now I am curious… maybe they are so subtle thatend up meaningless) but I was certainly not expecting release samples at different velocitiea in onboard piano engines (much more limited in sample memory).

vagfilm I don't recall what I was expecting, but if not new samples, at least different duration (a stacatto release might cut off the release sample earlier) or filtering of some kind; didn't hear any of that.

I just remember coming away with "oh that's it?" And thinking no wonder it's such a non-issue. I can just hear Mac muttering "I could've told you that," under his breath.

I will try to find some time to test this with mock midi files…

dore_m I mean, for years as a kid I wondered why I "ride the pedal", and then I forgot about it. But now I know I'm just controlling the brilliance, sustain, and preventing "muddiness" unconsciously.

That's one of the reasons, why detailed pedaling instructions are usually not written into the score. There have been written entire books on pedaling, as it is expected that you know how to use the pedal.

Digital pianos aim at the goal that application of the correct technique doesn't sound wrong through built-in amplification, however they don't provide any means to practice that technique.

Recorded sample libraries cannot sound wrong. There is no need control anything, which has already been picked up by microphones, selected for perfection and stored away to be triggered through MIDI. Things like riding the pedal simply aren't necessary.

QuasiUnaFantasia it? When a physical key is let go in an acoustic, gravity pulls, and the damper drops, where does velocity play a part?

MIDI was not designed to capture the events of digital pianos but to interconnect synthesizers and computers. In a synth, the note-off velocity can be used to control the Release component of the ADSR envelope.

In an acoustic piano the velocity of releasing a key has impact on how fast the key damper touches the strings and therefore on the sustain. This can have a noticeable impact on sound, especially in slower pieces that do not rely on the sustain pedal. Have a listen at András Schiff playing the WTC at the BBC Proms 2018 for several examples. But one could argue that this effect is minor when considering a wider repertoire.

In a digital piano, the note-off velocity can be a constant value or represent the release velocity at a given point. This feature requires measuring the time difference between a pair of key sensors - and this is the reason why DPs that rely exclusively on hammer sensors cannot implement it. As @CyberGene pointed out, some implementations, notably Yamaha's, represent the key release not as a single event but as multiple events. This better simulates the key damper mechanism of the acoustic instrument. Improvements to how series of note-off events are managed is actually one of updates to the MIDI 2.0 spec. In any case, the primary use case of note-off velocity is to control the Release of synth sounds.