Playback of slurred notes

Unless you have some ideas on implementing these together, I would say these are *not* there same thing.

16887 is about the *length* of notes slurred versus not. But what I am talking about here isn't note length at all, it's the *attack* of each note. On wind instruments, non-slurred notes are produce using the tongue for each note: "ta-ta-ta-ta". Surred notes get a tongue on the first note only: "ta-ah-ah-ah". I believe the way this is implemented for at least some synths is to turn on the legato cotnroller after the first note on, which should cause the synth to suppress the "attack" portion of subsequent notes. Other implementations might actually switch to do a different set of samples for the slurred notes. I have no idea if fluidsynth supports either of these "out of the box" or if the soundfonts themselves need to be set up to support it, or if it's implemented at all, but that's how I've seen or heard of it working in other programs.

This isn't just winds, btw - it's also for strings. Non-slurred notes are normally played with a change of bow direction for each note: "down-up-down-up". You can mark the to all be played with a down bow, but that means a full stroke for each note, which actually makes them more detached as you have to take the bow of the string between each note to reset: "down (up) down (up) down (up) down". Whereas slurred notes would be played on a single bow motion: "doooowwwwnnnn". The effect on the sound is similar to what it is for winds: less attack, leading to smoother sound. There is not necessarily any difference in note length between down-up-down-up and dooowwwwnnnn; it's all about the attack.

Note length is good to get right too of course, but it's really separate from the note attack issue. This might sound like a subtle distinction, but it's really not. Hearing an eighth or sixteenth note line in winds played tongued sounds very unrealistic. my opinion remains as stated above: this is probably the single biggest issue remaining with MuseScore's playback, I'd rank this above playback of rit, for instance, since that at least can be faked. On the other hand, I have no idea is actually overlapping the notes on playback would create a convincing simulation of slurring, by "covering up" the attack a bit, and maybe that's what you had in mind?

BTW, a complicating aspect of this is that while slur marks in wind music virtually always are supposed to be interpreted as "don't tongue", slur marks in string music are also sometimes used to indicate phrasing in more or less the same way as in piano music. A string player can only play a small handful of notes on a single bow. So if you see a slur across two or three notes, it's a bowing instruction. But if you see a slur marked over two entire measures, you need to realize it isn't really telling you about bowing - you'll still need to bow notes individually or in smaller groups - but it is really telling you the same thing it is telling a pianist, which is more the subtle stuff about note length and the final note of the phrase in particular. I think what this means is that it should be possible for the user to turn off the legato controller effect for any given slur.

I think the way you are proposing implementing slur playback for piano is very sound (although as noted elsewhere, I actually think this should be defeated by default - only users who ask to have non-slurred notes played non-legato should get that effect).

Unfortunately, this issue isn't about piano - it is about winds. Winds are totally different. Slurring for winds isn't a question of note length; it is a question of attack. Slurred notes should have no attack (or only a minimal one) - the tongue is not used.

My understanding is that sending a "legato controller on" message is supposed to achieve this in a synth. But I tried hacking this up once in rendermidi.cpp (replacing pedal on events with legato on) and nothing happened, so I gave up.

Piano doesn’t use single note dynamics, so it responds to velocity the old-fashioned way that was indeed common back in 1998 - it controls both the sample used (and hence the attack) as well as the volume. If the particular soundfont you are using does not sample ppp at differently than fff, you can simply try different soundfont that does.

In theory, those instruments that do support SND could take slur playback the way you describe - using a very low velocity to get the subtlest attack, then immediately ramping up the volume. Realistically, thats not all there is to slur on most instruments, I’m hoping to see a more sophisticated implementation for MuseScore 4.

The default soundfont works well enough for me. I haven't really experimented with different soundfonts that much, so far I haven't found any general purpose ones that I like better overall than the default (well, except maybe the HQ variant), and managing different soundfonts for different instruments is more than I have time/patience for. Of course, as a developer, I'm constantly messing with my settings to try to reproduce various user-reported issues, and then wanting to easily get back to defaults, so I tend to just prefer defaults over anything that takes work to set up.