Re the score, I'm using the Dover edition, which is a reprint of a Russian edition which lists an ossia for the most obvious textual deviation on my part (i.e. playing the climax with tremolandi and not repeated chords). Re other differences, the chord I play at 4.37 is Wagner's harmony: Liszt's is for some unexplained reason different. There are a few places where I have a couple of additions and/or alterations in the left hand: one bass octave is an octave lower (iirc) for emphasis and there are a few very small changes which were suggested by one of my teachers for consistency with the original orchestral score. (Though I don't like Horowitz's recording, I'm pretty sure he makes those changes also). The introductory four bars - well there was quite a discussion over this between me and my teachers! They are nothing to do with Mild und leise
itself, and I think that, with them, Liszt has written out an improvisatory introduction using motifs from the opera itself: of course it was the fashion in the 19th century to preface pieces in this manner, and they serve the dramatic and practical purpose of, during live performance, drawing the audience's attention and not "starting unheard". There was some consideration given to not including these bars: in the end I felt it safest to keep them (of all the recordings I've ever heard of this piece, only two pianists, Michel Dalberto and Lazar Berman, have omitted the introduction). What I did do in the intro is to distort the rhythm a bit in bar two: everyone seems to play the appogiatura onto the B natural as a very ugly, un-Wagnerian slide.
Other details that I can think of: I have a small misreading on page 2 (didn't notice it at the time, my engineer picked it up during editing but it was too late by then); also my opening of the aria I've really pushed what is acceptable with durations of the long notes. (The idea is to give a feeling of organic growth through the introduction and to give the impression of the music expanding around 1.00 to 1.15). There are a few other things which are very much personal decisions evolved over time; I do hope I've not grossly mangled the score (and also that there aren't small details wrong out of habit - after years of playing a piece they can creep up on you) but deep down I really don't think so! The dynamic return to pp at 5.10 isn't marked in the score, but is (I'm reliably informed!) performance tradition, to avoid the difficulties of the piano doing a sustained cresc over such a long period of time.
On a practical basis for your future performance, firstly I believe (as I think I said before) that keeping the tremolandi quiet is the single hardest thing in the piece. Playing them with a 'bebung' touch so that the keys never fully return helps. One of my teachers gave me a handy tip: if practicing on a grand, make sure it's open so you can see the hammers, and watch them as you play the tremolandi. If the hammers are jumping about a lot, the tremolandi aren't under control, but if they are just moving a little bit up and down, things will be going well!. Secondly I would suggest this is a piece where paying close attention to voicing will yield major dividends. The section which begins pp over a B major 64 chord (from 3.12): I practiced this so much and still didn't get it quite right! I think the top (soprano) line should float ethereally over the chordal accompaniment in the rh - in my score the rh is arpeggiated on the beats but I have seen editions where it's not. I practiced this section, for effect, playing the rh top stave as a lh and rh duet: lh chords ppp and soprano pp, using left before right to emphasise the separation (I think the arpeggiation marking justifies this - to be precise I was playing the lh chord in unison and viewing it as an acciaccatura onto the soprano melody note). Very difficult to emulate this effect when playing the top stave rh alone.
Thanks for listening, and good luck with your performance!