Velocity 1 = no sound

Thanks Portable hole, that is interesting because..

Joe you're 100% correct and yet there's some un-charted behaviour seeming to be in play too. In spirit of adventure and fun (and maybe to benefit Modartt's ongoing efforts in improving playability)..

If we set the velocity curve as you set out Joe, [1,1 and 127,1] then swipe a mouse across the virtual keys, you'll note that the faster you swipe, the more notes are dropped with no audio, even though red keys follow the cursor.

To prove that's not a mouse (x poll data points per ms etc.)..

If you next set the velocity to ~middle like [0,60 and 127,60] and repeat, you'll see the similar red keys following cursor but also hear more notes.

After beginning experiment and going incrementaly higher, the magic number seems to be velocity 9 where noticeably more notes are included.

I'm wondering if Modartt does discard a % of low velo notes, at least those with no 'note off' data? (to help playability). I do know I feel playability has improved so greatly in recent years but never notice this before

I imagine it varies from physical model to model to match the required velocity of the given instrument. I find the velocity response streets ahead of the others I have tried.
Mossy's comment mirrors what I would expect - the OP should consider that pianos are percussive instruments. 0-127 represents an extremely wide range, far beyond the sensitivity of the human hand while playing music (you couldn't even be expected to reliably or consistently aim for a precise value hitting just one key in a non musical experiment). IOW plenty of redundant data space if needs be to replicate realistic dynamic attack rather than making a sound at the lowest note on value for the sake of it. 

I noted that if you pressed so lightly as to not get a note on sound you would still get a note off sound, as per lifting fingers from the keys resonating through the body of the instrument, impressive attention to detail right there.

Believe you are 100% correct Joe about OP's video - the silent note has a large enough space to be unaffected by the inhuman ppp.p(recurring) glissando effect. I'm sure there's something relating to the keyboard, no note-off being fired, dynamics and internals of Pianoteq which produce that particular result.

This gives me the warmest laugh I've had in a while..

I can't help but imagine a VJ feverishly working on this skill for a Youtube super-duper glissando squared! 'how to' video

Pianoteq really does take human input into account so well - it is my 'next best thing' to a real piano for learning or improving, let alone recording 'the piano'. Its evolution is quite a thing to behold and I see it as a technical/scientific and artistic advancement of great historic importance in the evolution of the instrument - as well as the modellings other impacts in coming time.

Interesting thoughts Mossy and Key Fumbler. It's always fascinating to consider those things. My fastest limits of physical playing can't seem to break anything about the illusion of reality for me. Playing in real time I've never noticed dropped notes but my keyboard does send note-off data which the video shows as absent. Staccato behaviour at human speed is certainly one thing I've observed particularly noticeable improvement in Pianoteq over time.

Wondering how Portable Hole is getting on with any further experiments also.

In the MIDI standard, a Note On event with velocity 0 is to be interpreted as a Note Off (instantaneous). So I guess it is sensible to interpret a Note On with velocity 1 as no sound for a given duration.

On a real piano it is possible to strike slowly a key so that the hammer does not reach the string. If the key is held while other notes are played, the silent key will sound by sympathetic resonance. Pianoteq implements correctly this behaviour so I guess this is one possible reason. If you don't like it, simply start your velocity curve at (0,1). I also suppose this was decided because not many keyboards go down to 1 reliably.

If there is a better reason, maybe Philippe could explain.

Perhaps a sound is being produced by Pianoteq even at velocity 1 (assuming the velocity curve has not been altered such that 1 velocity is interpreted as 0), but is just too faint (too low in electrical amplitude) for the physical components of the sound system, the headphones the speakers or the ears to register.

At any rate, it seems that a simple adjustment of the velocity curve in Pianoteq would resolve the issue. In some cases however, a performer might not want extremely low velocity readings to trigger any sound at all, because in modern triple-sensor keyboards, that might trigger a sound even when a note is barely brushed or touched in passing by a finger, and the spurious, unintended note-sounds might become very frustrating.