Differences between versions 1.0 and 1.1

"One model that occurs to me is to have a separate program that reads MueScore files - or better yet, MusicXML files - and then provides an interface geaed specifically toward playback"

That sounds like a MIDI sequencer. You load a MIDI file and you can freely change it, soundwise but also by shifting notes etc... to make a "performance" out of a score.

Yes, I agree, a getting sound good enough for demo purposes is one of the useful things about computer-generated playback. But I would claim it already *is* good enough for that purpose. Again, if you have *specific* ideas for improvement - ideas that would integrate well within the model of being notation program first and foremost - feel free to submit them. In separate threads, and/or in the issue tracker if the suggestion is especially well thought through.

As for people who cannot read or write music - sure, obviously, they exist. But MuseScore is not intended for them. Just as there are people who can neither read nor write written language, and Word is not intended for them. And there are people who don't know how to design buildings, and AutoCAD is not for them. I fail to see how this has anything to do with a discussion of the purpose of MuseScore. MuseScore is a program with a purpose: to produce sheet music. You use the program if you want to achieve that goal, you use a different program if you have a different goal. The existence of people with different goals doesn't change the purpose of MuseScore.

Clearly there are many valid reasons which make the case that playback in MuseScore should be further improved. In fact Werner already implemented many improvements in the trunk, like the ability to use more than just one soundfont for playback. But as MuseScore is still very far of being a feature complete project, and the fact that there are limited resources on the level of core software development, we put notation features before playback features. That being said, we'd love to speed up the development dough. Thats why we actively travel the world and evangelize MuseScore to fellow developers. Fingers crossed ;)

Agreed; nice writeup.

I would observe there are really two fairly different issues being discussed here. The first is the use of MuseScore as a MIDI editor of sorts - which I take to mean, taking an existing MIDI file that may or may not have come from MuseScore and tweaking it in some way then saving as MIDI again. The second and really mostly unrelated issue is improving the playback capability of MuseScore - the extent to which a score that was created from within MuseScore will play back convincingly and I guess oh by way export a MIDI file that represents that playback, although I'd also expect the results of playing that MIDI file in a generic MIDI player to potentially be disappointing.

To me, the former - direct MIDI editing - is the issue that is not worth spending a lot of time on. There are too many technical imitations due to the nature of MIDI and the need to imort and export and lose information both ways. It's just too far outside the scope of what MuseScore is intended for and optimized to do well to expect much in this department, it seems to me. I can see a few relatively simple things that would make a pretty nig difference in practice, though - mostly, more control over the quantization of notes on import, preserving of more of the MIDI performance data by interpreting it and encoding it within the existing framework as note properties, etc. If a facility to attach arbitrary MIDI messages to staff text were added, so much the better (looks like 2.0 has some new stuff in this area?). MuseScore would never replace a MIDI sequencer for the things a MIDI sequemcer and the MIDI format is optimized to handle well, but I can certainly see these and other tweaks making it more useful for the relatively small number of users who would be doing this sort of stuff.

But the second issue I mentioned above - improving playback as a separate thing entirely from MIDI editing - is something that to me fits in well with what can reasonably be expected of a notation program and what many would find useful. And 2.0 looks to be doing a decent amount of that already, with playback support for more markings, ability to change instruments for a staff md-score wihout the need to predefine an instrument in instrumemts.xml, etc. Realistically, though, programs like Finale have a pretty huge headstart here. From a purely pragmatic standpoint, I suspect the programming resources required to close this gap would be such that it would pretty much bring other development to a halt for an unacceptably long period of time.

Which is why I am starting to get kind of excited by my idea for a playback engine as a separate program for someone else (other than the core MuseScore team) to work on. I could see it using, for example, that "sinfonia"-whatever sample collection that people seem so excited by, as opposed to being limited to using the soundfont format via fluidsynth. And providing all the sorts of controls I mentioned earlier - ways of setting default behavior of various markings, also override these defaults for particular selections, etc. Also easy ways of "nudging" the timing of notes via a direct interface of some kind, as opposed to fiddling with numbers in a dialog as MuseScore currently offers. If there were a way to somehow integrate this with MuseScore so it looks seamless from a UI perspective, great, but I also think that it would be a mistake to assume that all the data you create using this program would need to be saved in MSCZ format to then be further edited using MuseScore proper. I mean, that would be wonderful in theory, but that would require the sort of architectural changes to MuseScore I'd be wanting to avoid with this proposal. I think we'd be better off in the long run if we accepted this, like MIDI export, being essentially a one-way street.