The vowel you most need is the one in the second syllable of water and the first of about. It is called the schwa, it is the most frequent vowel sound in English, it has no letter of its own, and it is what almost every unstressed syllable turns into.
Which means it is not one of the twenty to be learned alongside the others. It is the destination the other nineteen collapse into whenever they are not carrying the beat, and a learner who never produces it is producing full vowels in every syllable of every word.
The eight vowel contrasts that matter most
- ship and sheep. The first is short and relaxed, the second long and tense, and they are different vowels rather than the same one held longer.
- full and fool. The same contrast in the back of the mouth, and equally invisible to speakers of Spanish, Italian, Greek and Japanese.
- cat and cut. Two sounds that many learners collapse into one, and which separate an enormous number of everyday words.
- bed and bad. The second is longer and more open, and confusing them is the commonest error for German and Persian speakers.
- the schwa, the unstressed neutral vowel, which is more frequent than any other sound in English and appears in no spelling of its own.
- about has two schwas in ordinary speech, first and last, which is why it sounds nothing like its spelling.
- caught and cot. Distinct in Britain and merged for most Americans, which is a genuine dialect difference rather than an error.
- bird, whose vowel exists in Britain without the r being pronounced at all, and with it in most of America.
Why the twenty are crowded rather than numerous
Twenty is not a large number of things to learn. What makes twenty vowels hard is that they are packed into a mouth with no gaps in it. Your mouth does not have twenty positions; it has a continuous region, and English cuts that region into twenty overlapping neighbourhoods where most languages cut it into five or six wider ones.
So a vowel is not a target you hit but a region you land in, and two English vowels are frequently two halves of one region your ear already owns as a single place. The distance between ship and sheep is not a distance your language ever asked you to measure, and it is smaller than the distance between two allowable pronunciations of a single vowel in most languages.
It is also not a matter of length alone, which is the usual first assumption. The two differ in where the tongue sits as well as in how long the sound lasts, and a learner who reproduces only the length produces a long version of the wrong vowel, which is heard as the wrong word said slowly.
Where it stops: giving every syllable a full vowel
This is the stall, and it comes from good advice applied one level too far. Told to speak clearly, a learner pronounces every syllable fully, so photograph comes out as three equally weighted syllables with three real vowels in them. Every sound in it is correct. The word is much harder to recognise than it would have been said carelessly.
It stalls here because the feedback runs backwards. When somebody does not follow you, the instinct is to say it again more clearly, which means more fully, which flattens the word further. The learner concludes their vowels are not accurate enough and works on accuracy, and accuracy was never the problem.
The repair is to say less rather than more. In an unstressed syllable, aim for the schwa and let it be short and colourless. It feels like sloppiness from the inside and it is what the listener is waiting for.
Hearing the spread rather than one idealised version
Forvo is a free archive of words pronounced by ordinary speakers, usually several per word, and its value for vowels is exactly that plurality. A coursebook gives you one studio recording of bought and you copy it; Forvo gives you six speakers and the vowel they share is the vowel, while everything they differ on is the part that does not matter. That distinction is invisible from a single recording.
It replaced the single voice on a coursebook CD, which was one person's production presented as the sound. The limitation is that recordings carry a country label at best, so you may be copying somebody whose vowel system differs from the one you are aiming at, and for the crowded pairs above that difference is larger than the difference you are trying to learn.
What it looks like when it goes wrong
The recognisable failure is a word that is understood at half speed and lost at full speed. The vowels are right in careful speech, the schwa is missing, and once the sentence speeds up the listener has nothing to tell the stressed syllables from the rest.
The second one is a merged pair, where ship and sheep are being produced as one sound and heard as one sound, and nothing in the conversation reveals it because context repairs almost every instance. The way to force the split is to remove the context, which is ship and sheep: training your ear on one sound at a time. The symbols for all of this are on a symbol per sound, so an unfamiliar word becomes sayable, and both sit under being understood the first time you say it.
What decides which syllable gets the full vowel and which gets the schwa is not the vowel at all, and it is not a property of the sound. It is a property of the word, and it is the next thing to settle: which syllable to hit, and how the wrong one changes the word.