Visemes

[Fontanellean]

Well, the problem with Voynichese on the level of glyphs (or glyph clusters) is that their position and company is too predictable. 

Deaf people are aware of word boundaries (nouns, names, verbs... are understood as units of meaning). So we assume spaces are unaffected. Now take something like the vowel "a" in Latin, which I assume would be represented separately. We need that to appear word-initially (ad, ab, ac, atque...), medially (infinite examples) and word-finally (infinite examples). Dream up a position in the word (or sequence of sounds), and Latin has frequent examples of words with the a-sound in that position. Changing the underlying mechanism to lip reading does not change the fact that Voynichese does not offer this kind of flexibility by a long shot.

Now, you could say the deaf person makes separate glyphs for "am", "an", "al" etc because people's lips may look different depending on the neighboring sounds. But then you'd need an awful lot of glyphs and still a lot of flexibility. 

So in short, your proposal doesn't seem to solve any problems but in fact might make things worse.

I suppose another defeater is the low abundance of two-glyph words on certain pages.