V and W: why apps keep telling me I said the other one

If it's happening consistently and in one direction — always v-for-w or always w-for-v, not randomly — the app is very likely right, and the fix is a specific articulation change rather than "speaking more clearly."

30-second version: English /v/ is made with the top teeth touching the bottom lip, producing friction. /w/ is made with both lips rounded and no teeth involved at all — it's a glide, not a friction sound. Several languages (Hindi and Urdu among them) have one sound in between the two, so a learner's /v/ and /w/ can genuinely land in the same acoustic space, which is exactly what a phoneme recognizer is built to notice.

Why it happens

Hindi and Urdu use a single consonant, transcribed /ʋ/, that sits acoustically between English /v/ and /w/ — some lip-teeth approach, but usually not enough contact to make the friction that defines /v/, and some rounding, but not the full glide of /w/. That single sound comfortably does the job of two English phonemes at home, so when a speaker maps it onto English, the two target slots collapse into one. It's a genuine sound-system mismatch, not a matter of trying harder.

What makes this pair unusually easy to test for objectively is that both directions are common: some speakers systematically produce /w/ where /v/ belongs ("very" → "wery"), others hypercorrect once they know about the issue and start producing /v/ where /w/ belongs ("water" → "vater"). They need opposite drills, so knowing which direction you go matters more than knowing that you "have a v/w problem."

How sayit catches it

The two sounds differ physically in a way an acoustic model can measure directly: /v/ has continuous frication noise from the lip-teeth constriction; /w/ has none, just a smooth formant glide into the vowel. sayit's phoneme layer reads that evidence from the waveform itself, independent of which word is more common in that sentence position. When you say "vet," the target is /v ɛ t/; if the acoustic evidence lines up with /w/ instead, the per-word breakdown shows a substitution at that slot — target /v/, produced /w/ — with the "sounded like" hint showing the real word it landed on if there is one, which for "vet"/"wet" there is.

Because the detection is symmetric, it catches the hypercorrection direction just as reliably as the original swap, which matters if you're past the first stage of fixing this.

Finding your direction first

Before drilling, find out which way you actually go — most learners are consistent in one direction and inconsistent in the other, and guessing wastes practice time on the wrong half of the problem.

Record: "We value very wide views on this." It alternates the two sounds five times in seven words, so one take usually shows a clear pattern in the breakdown.

The fix, in both directions

For /v/: top teeth rest lightly on the bottom lip and stay there through the sound — you should feel a buzz on your lip. Overshoot the contact during practice; it settles down once it's automatic.

For /w/: round the lips firmly, like starting to say "oo," and make sure the teeth never touch the lip at any point in the sound. If your teeth are anywhere near your lip on a /w/, that's the leak.

Drill the direct contrast at wine vs vine and vet vs wet, then check the workplace-frequency pairs — vary/wary, verse/worse — since those are the words that give the pattern away in a meeting rather than a drill. The full article on this pair goes deeper on the physics if you want it, and /phonemes/v and /phonemes/w have the articulation steps.

Tracking the fix

Re-record the same alternating sentence every few days rather than switching to a new one each time — consistency across a fixed sentence is easier to read than comparing different sentences. When the verdict on both the /v/ words and the /w/ words in that sentence stays good across three separate takes on different days, the contrast has generally moved from "drill success" to "habit," which is the distinction that actually matters.

Try it

Open sayit free — no install, no card — and read the trap sentence above. If one direction keeps getting flagged and the other doesn't, you now know which drill to actually run.