One great benefit of working with LilyPond is the superior quality of its out-of-the-box results. If you’re just creating a score for your own use you rarely have to intervene manually at all in its layout decisions. If you provide others with performance material (for example if you’re a composer) you may have to fix a few more relevant imperfections, but these usually don’t even affect the read- and usability of your score. Quite the contrary: through its well-balanced and weighted overall look, any default LilyPond score is really comfortable to read in real life, i.e. on the music stand. Only if you are going to publish a score you will have to go into typographical details.
Much of this superiority is due to the compiled approach LilyPond is taking: It doesn’t act like a specialised graphics program (which WYSIWYG programs are to some extent) but first reads and interprets the musical content, producing a strong internal representation of the score’s structure. Only then it starts to consider layout decisions and takes the time it needs to consider side effects etc. While this isn’t perfect yet (and maybe never will be) it is usually far superior to anything the competitors can achieve. The potential becomes especially obvious when the music gets complex, and I want to show you a quite spectacular example I encountered some time ago.
When I had to play Arnold Schoenberg’s fragment Gethsemane (which Schoenberg had sketched as the piano reduction of a to-be-written orchestral cantata) I had to reduce a section from four to two staves in order to be able to play it – resulting in a very complex piano score. When I had entered the music to LilyPond I was totally stunned by the result. While I wouldn’t immediately publish the score there wasn’t anything that disturbs reading the music. Please note that in this score there isn’t any single manual improvement of the engraving – all I did was assign the music to voice numbers (1-4), and LilyPond took care about the rest!
Well, I have to admit that you won’t always get this level of default perfection – this is actually an exceptional example (I think the secret lies in the fact that there are so few slurs in these measures). Nevertheless I’d be interested in seeing attempts from other programs, as long as they too don’t apply any improvements over simply entering the music. But how would I know they’ve been fair and didn’t secretly improve their typography manually? Interesting question, stay tuned …