Creating expressive MIDI performances

• Jul 16, 2019 - 06:40

Been using musescore for a few days and I'm really amazed by the quality of the software. Hats off to the team.

But the problem is I started this particular project with the goal of producing a highly tweaked MIDI file that sounds just like a live performance. In other words, to make this https://e.issuu.com/embed.html?identifier=c8v6se1wfpbp&embedType=script… sound like this https://open.spotify.com/track/5HxMisNV5qiXl0t7ORgQQP. As usual, there's a lot more going than what you see in the score.

In musescore, it seems there is not much ability to dial in performance contours, i.e. tuning a crescendo by tweaking nodes on a bezier curve. I know musescore's primary concern is notation, not synthesis. Any workflow tips with MIDI performance in mind? Any crucial features I should check out? Do folks end up doing these tweaks in other software? I realize it's maybe not a very common objective.

I'm a transcription novice in general, this is perhaps more of a question about MIDI transcription philosophy. It would be nice to arrive at a performance-quality MIDI by building up from raw notation and tweaking from there, but maybe it's simply not done that way.

Comments

• Jul 16, 2019 - 14:35

For the sort of really extensive tweaking you describe, what makes most sense to me is exporting to MIDI then editing further in a sequencer / DAW.

Reply

• Jul 16, 2019 - 17:09

Two things come to mind.
First is that each of us has a different idea of "sounds just like a live performance". So much so that it almost seems like a useless quest. What makes a file sound "real"? We each have our own answers. The only way to get really close is with a powerful computer, expensive software, and a few thousand dollars in sound libraries. And the know-how to make it all work. Which brings me to ...
Second is that the new version of MS is, as you say, amazing in that so many more expressive aspects are definable. Spend some time in the manual learning how the inspector works.
And your end goal is important. If you are transcribing a piano piece to be played by an orchestra, than that requires one approach. If you are after just a sound file, something different might be needed. Both require skill in orchestrating, and years of training. If your transcription is full of things real players can't do, it's not going to sound "real". If your transcription is full of instrument combinations and techniques that don't go together, it's not going to sound "real".
And spend some time checking out the various sound fonts available. Learn how to use and combine them. each has strong and weak points.
It all takes time. Have fun.

Reply

• Jul 16, 2019 - 17:21

This is all wise. There's no "make it sound like humans" feature. Increasing the believability of automatic performance is a never-ending task with countless byways and side-streets. Indeed, learn to use what's there, including the Piano Roll Editor (which finally became usable in MS3). Easy control of phrasing will be present (via plugin) in the next MS release.

Reply

• Jul 17, 2019 - 01:51

Just a few thoughts.
For me, "make it sound human", and "make it sound believable" are two different things. A group of professional musicians gets together and either they put on a concert and/or make a recording. Do they play every note perfectly? No. But that's not the point nor the goal. If they could, they would make every performance perfect. Why on earth would I want to take my rendition of some of my music and make it sound human? Why would I add timing mistakes and pitch problems to it? Real musicians don't try to do that. Besides, how often do you hear things like that in a quality recording produced in a studio? I go back to a time when there were plenty of junk recordings. Lousy mic placement or lousy mics or lousy recording techniques, and equipment. But the musicians still tried to play the best they could. It's what they do. They try to "make it sound believable".
But I would go one further and replace both those criteria with "make it sound musical". I know folks who have all the bells and whistles as far as hardware and software goes, and the skill to back it up. They write and produce great sounding stuff. They don't like to listen to notation rendered music. Can't blame them.
MuseScore is quickly coming to the point where "make it sound musical" is becoming a thing. I do wish that inspector definable tempo changes and ability to snap to at least a eight note in an empty measure were possible. OK that last one has nothing to do with sounding musical, but would sure speedup notation.
So, my new mantra is just to Make It Sound Musical.

Reply

Do you still have an unanswered question? Please log in first to post your question.