The Maximal Onset Principle states that where there is a choice over the syllable in which a consonant, in this case /v/, is to be placed, it goes into the onset rather than the coda, that is to say, at the beginning of the following syllable, rather than the end of the preceding syllable, but this applies only within certain phonotactic constraints. One such constraint is that a syllable may not end with a vowel such as /æ/, and this will almost certainly explain why you have seen it as /'græv.ɪ.ti/ in most dictionaries.
Syllable boundaries are of interest to the phonetician, but they don’t normally have as much bearing on actual pronunciation as the stress, which in this case is correctly shown as being on the first syllable.
So which pronunciation is standard for the [ʊ] sound? Rounded or unrounded?
Certainly there is some rounding, but because roundedness is not phonemic in this position, there is also considerable variation in how much of it actually occurs in any given word and speaker.
For example, you will find that it is generally somewhat more rounded in pull and full than it is in put and foot respectively. That’s because having an r or an l right next to it rounds it off a bit — which is why it is a bit more rounded in root and rook than it is foot or cook. Same with rookie versus cookie, where the first version is a bit more rounded than the second. And of course, a w helps: compare how wool is even more rounded than full, and also moreso that wood.
I believe English has no words with [ʊw], as that seems redundant. However, it can occur in phrases, especially in some dialects, where something like I knew it full-well may approach that.
However, it is still perceived as the very same phoneme in all those words and cases I’ve just listed above.
Correction — or not
I said that I thought English had no words with [ʊw] in them. And at the end of the day, I still believe that. However, I have discovered that grepping the OED yields the apparent existence-proof counterexample of Rauwiloid, which means:
A proprietary name for a hypotensive preparation containing a number of alkaloids extracted from Rauvolfia serpentina.
You also have compound words whose first element ends in [aʊ] (rather than [aw], as it is sometimes spelled) connecting to something that begins with [w], and which have in effect a “double w” in them, you expand the list to include such things as:
bow-wow, powwow, skeow-ways, wow-wow
Finally, if you consider the sound in words like no and micro to be
an [oʊ] diphthong rather than [ow], then you get all these, most of which were originally compounds of some sort:
froward, frowardly, frowardness, glow-worm, Holloway,
hollowwort, Howeitat, Khowar, meadow-wink, microwave, microweld,
Moldo-Wallachian, nowise, Oldowan, Parowax, powan,
shalloway, slow-worm, swallowwort, werowance,
yellow-wood, yeowoman.
For example, yeowoman theoretically yields /ˈjoʊwʊmən/, at least in North America. Still, there is a reasonably convincing argument to be made that that one is better written as simply /ˈjowʊmən/.
Slightly less uncommon is nowise, which is a compound of one word ending in a diphthong connected to another starting with a triphthong, so /ˈnoʊˌwaɪz/.
But I am still highly dubious of the existence of [ʊw], because I think it fuses into the semi-consonantal glide, [w]. After all, nowise and no eyes are homophonic, so I think this idea of [ʊw] is very hard to justify, and so I stand by my initial statement.
Even towel is usually pronounced with just one syllable, /taʊl/, thereby rhyming with cowl /kaʊl/. Even with folks who work very hard to put two syllables into that, with /ˈtaʊ.wəl/, I submit that you could write that /ˈtawːəl/ and avoid the whole controversy of whether a semi-vowel/semi-consonant/off-glide is really /ʊ/ or really /w/. However you write it, it seems like the same sound to me, such that bisyllabic towel just has a geminate [w]: /ˈtaw.wəl/.
Best Answer
Not all English speakers pronounce all words with the same number of syllables.
However there are general agreements as to how many syllables any word has. This is certainly true in singing.
In traditional music, one syllable will be sung on one note. Of course, if there are more syllables than notes, one "cheats" and adds notes. More notes than syllables, then one can "cheat: by distorting the word to created new syllables. Many songs have an even syllable to note equation. The "cheating" usually happens when there is an attempt to fit too many or too few words to an established tune,
In the well known song "Mary Had a Little Lamb", there are exactly the same number of syllables as notes in the tune. I have never noted any problems with this song as to syllabification.
Often persons who are not native speakers of English and speak English with an "accent" exhibit no such "accent" when singing English songs. Native and non-native speakers follow the same syllables with the same stress dictated by the tune. If one wished to assure the convention as to number of syllables is followed, one could do worse than to practice singing single syllable words on a single note
Almost always, the number of syllables in an English word is determined by the number of sounds (vowels) in the word; not the number of written vowels, but the number of distinct vowel sounds.
The second vowel in "little" is not written. Tl could not be pronounced without the understood vowel.
Etymology would have little to do with syllabification, unless one were trying to pronounce a current English word in the manner of some previous era. That does not occur very often.
I'm not sure this is a very serious issue for the most part. Probably few care if a syllable is added or subtracted in speaking as long as the meaning of the word is clear. Unless one is singing.