The trick with voice leading is that the concept of voices is meant to be somewhat more abstract than most people realise.
It is no accident that voice leading is traditionally taught in four voices. One of these is the bass voice, which is unique, and obeys its own set of rules, because of its need to frequently sing/play the root of the chord. The remaining three voices represent each of the three tones in a triad, which must each move in certain ways relative to each other. In traditional voice leading instruction, these voices are treated as literal voices, that will sing actual melodies made up of the pitches that voice leading dictates (note that there are often multiple options).
As an example, when moving from a triad to another triad whose root is up a fourth, ignoring the bass line, the traditional patterns are: Common tone (the root of the first chord is kept to be the fifth of the second chord, and the third and fifth of the first chord both step up to become the root and third of the new chord, respectively); Contrary to the bass (all three notes move downward to the nearest chord tone - the third must actually skip down to become the fifth of the next chord); Tertian Leap (a combination of the last two: the root of the first chord is kept common to become the root of the new chord, but the fifth falls to become the root of the new chord, and so the third must leap all the way to third of the new chord in order that the new chord has a third - in traditional voice leading, the chord must always have a third); and Triple Root (again, a combination of the first two, where the third rises and the fifth falls, both to become the root of the new chord, but the root skips down to become the third of the new chord - there is no fifth, which - unlike the third - can sometimes be omitted).
There are a couple of points to this. First, these patterns are just ready-made combinations that happen to avoid the true voice leading errors (of which there are only a few). No parallel fifths or octaves (if both chords are in root position, that means that in the upper voices, the root should never move to the root, the fifth should never move to the fifth), the chord must always have a root and a third, and should usually have a fifth.
Second, the voice leading of a voice partly determines and is partly dependent upon the other voices. For instance, if the melody moves from the fifth of the chord up to the third of the next chord, the other two voices will most likely move as in the common tone patter (though atypical voice leading is possible if no rules are broken - however, this often makes proper voice leading in subsequent chords more difficult, which is part of why the patterns are used).
Most importantly, though, the voices aren't real. They are artificial constructs that are created by the patterns. It might be more accurate to call them "group chord-tone tendencies". It doesn't really matter if there are voices, or how many there are. What matters is how to move the various combinations of notes. The guitar is a perfect example. In spite of preceding comments, the guitar is actually surprising good at voice leading - as long as you don't care about the number of voices (which are all artificial anyway). The traditional chord shapes mostly don't cause parallel motion (though barre chords do). Suppose you were going to move from an E major chord to an A major chord. In the E major chord, the root (E) is on both the fourth string and the first string, and in both cases, it stays common to become the fifth of the A major chord. It is important to keep in mind that these are not two different voices, they are two instances of the same voice. The third of the E major chord (G#) is on the third string, and it rises to become the root of the A major chord. The fifths of the E major chord is on the second string, and it rises to become the third (C#) of the A major chord. The same tone is also on the fifth string; once again, it is just a second instance of the same voice (not another voice), but this time it doesn't follow its partner,, it disappears, and the root from the sixth string moves up to the fifth string to become the new root (this is a bass line, and follows different rules). There is absolutely nothing wrong with the duplicated voice or with the voice that disappears. All that matters is that the notes themselves move according to one of the acceptable patterns.
To be fair, I should say that a single note that is doubled might be able to move in different directions. For example, after the triple root pattern, the new chord has three roots (one is in the bass) and no fifth. In this case, the two upper roots must move differently in order to avoid parallel octaves and in order to get a complete chord in the next chord. In the case of the guitar, a true fifth voice may even show up, where multiple voice leading patterns are used at the same time, as long as there are no parallel fifths. But a root still will not move to another root (other than the bass note or in inversions), and the fifth still will not move to the fifth.
The guitar makes another useful point. These chords need not be strummed all together. They are often arpeggiated, meaning that they are played one after another, melodically. When you do this, you are not creating a melody. The notes themselves remain as representatives of the imaginary voices, and even though they might occur on opposite ends measures, several beats apart, they should still move according to these patterns. Suppose you arpeggiate the E chord from the bottom up to the top, and then the A chord from the top down to the bottom. On the page, it looks like a melody that skips its way up and then back down, but in reality, it is multiple voices that are still connected to each other across the distance, as though all the notes were being played together. And remember, the voices can enter and leave at any time, they can be doubled or lose their doubles at any time. And if the notes are far enough apart, the voice leading doesn't really even matter.
When an actual melody does contain a leap, in a way, it is like the melody is moving from one "voice" to another - remember, these voices are abstract, not literal - and it takes on the voice leading tendency of the new voice it has moved into. At any given moment, as a chord is about to change, the melody can be assigned to one of these abstract voices, and must move according to the rules; if it was previously on the fifth of the chord, it can move to the root or third of the new chord, but traditionally not to the fifth (it does sometimes happen in other styles of music, but this is simply because the rule about parallel fifths is style-specific, and not universal). Once this note moves, any other notes that may happen to move at the same time, or nearly the same time, must move in ways that fit the patterns - or drop out. If you pick "any random bar from Haydn", you should be able to see that this is so.
In the special case of guitar with, say, power chords, or bar chords moving in large leaps in parallel, we still have a kid of voice leading, it is just a different kind. A classical theorist might call it poor voice leading (lots of parallel fifths and octaves), but that is stylistically okay in most modern popular music. More sophisticated theorists may suggest that certain voices are simply shifted up an octave and then move correctly; this is a bit of a conceit, but it suggests another way of thinking about voice leading. Imagine a kind of graph paper or template paper that you could write on. Suppose you have a G chord. On your template paper, you would have faint watermarks of all the possible notes of the G chord (GBD) in all possible octaves lightly highlighted, with faint arrows showing all the places they could go to when the chord changes. With the understanding that anywhere a note can go in its own octave it can also go in other octaves (and so not showing all these octave possibilities), the paper would show a repetitive, wallpaper-like pattern all the way down as each voice reappears in subsequent octaves. In a way, this wallpaper is implied underneath all music, and the melody, rather like finding a way out through a maze, just traces out some of those arrows.
In summary, voice leading does not apply to true voices, but rather to notes that represent abstract voices. In a piano part that uses two hands, even though it looks like there are only two voices, in an abstract way, there are four voices, and the piano parts weave their way in and out of the various abstract voices, and when the chords change, the notes must move according to the patterns for whatever voice they happen to be in at that moment. Consider Bach's Cello suites, for instance. Only one instrument, usually only one note at a time, so it looks like one melody, but really, it is jumping around between abstract voices, and they all move as though they were a choir (I promise you this is true, I have spent a great deal of time tracing them out to verify).
