Beautiful new score, midi rendered with lmms

I posted a few weeks ago about buggy copy+paste operations in Musescore 1.3 that were standing in the way of creating acceptable guitar bariolage playback. (Bariolage = the same note played successively on two different strings.)

Faced with a barrier so daunting I was inclined to set the project aside, I decided to approach the problem of bariolage from a different angle by exploring the capabilities a sampler different from the one I normally use, LinuxSampler.

My idea was that if I had finer control over the sound of my instrumental tracks, I could create the bariolage effect outside of Musescore.

Long and short of it, I connected Musescore to 'lmms' (Linux Multimedia Studio) and used its sf2 player to render my score. The results exceeded all expectations. I’ve posted the audio+score on YouTube at http://youtu.be/qGaalpwOE9A along with an audio+lyrics version at http://youtu.be/wLbMi2m7Rm4. Well worth checking out.

Although lmms can only handle .sf2 soundfonts, as a sampler it has one mighty advantage over LinuxSampler: you can apply equalization and other effects separately to every track. What this means is that the timbre of each instrument can be individually tailored—on-the-fly, if necessary, during the final mix. My piece, a song called Remember, is scored for a solo cello supported by clarinets, bassoons, high strings (6 violins I, 4 violins II), arco bass, and guitar. It’s a difficult combo to work with, and would never have come off successfully without the ability to add glassiness to the high strings while maintaining warmth in the middle register, bringing out the “chestiness” of the bass in its upper register, or achieving decent bariolage in the guitar part. (Not to mention getting the clarinets to sound right in both the chalumeau and clarino registers.)

lmms isn’t perfect. There are issues with single-staff polyphony (if a note is being held in one voice on the staff, the same note is dropped every time the second voice requires it). There are occasional and inexplicable dropped notes under other circumstances—very few, but enough to have required three takes for my final mix to get the missing notes.

In addition, one is limited to 16 channels, which means firing up a second instance of lmms if more are needed (say, for a full orchestra). LinuxSampler lets you set the port and channel for each track, but lmms does not. Lastly, one is restricted to .sf2 soundfonts because there’s no .sfz or .gig player.

I’ve posted to the lmms forum about the first issue (single-staff polyphony), but as yet, no one has responded. This has me worried about the state of lmms’ development, since the problem is not trivial and ought, to my way of thinking, to have raised a red flag with the developers. With that bug fixed, lmms would be the ideal sampler for MuseScore. It has a beautiful, smooth interface, provides a huge FX board that can be used as a mixer (solves Musescore’s crescendo/diminuendo limitations), and allows for tailoring of timbres while permitting dynamics, articulations, and tempi to be controlled from the Musescore score, which is the logical place from which to perform such operations.