Which of the 44 English sounds are my problem sounds?

By recording real sentences and letting a per-phoneme heatmap, built directly from your own speech history rather than a generic checklist, show you where the color-coded flags actually cluster — because guessing which of the 44 sounds are yours wastes time on sounds that were never a problem, and drilling all 44 equally wastes even more.

30-second version: Every phone in every word you record gets compared against its target and colored by how it scored — a clean match reads as confident and correct, a substitution or deletion reads as flagged. Do this across enough recordings and a pattern emerges: most learners' actual problem sounds cluster to three to six specific phonemes out of the 44, not a scattered mess across the whole chart. That short list, built from your own evidence rather than a general list of "hard sounds for your first language," is the one worth spending practice time on.

Why a generic "hardest sounds" list isn't the same as your list

General guides to the hardest English sounds — this site has one — are genuinely useful for understanding why certain sounds are widely difficult, tied to broad patterns across first-language backgrounds. But "widely difficult for speakers of your first language" and "actually difficult for you personally" aren't the same list. Plenty of learners have already, through years of exposure, fixed one or two of the sounds their language background typically struggles with, while still carrying others most guides wouldn't flag as unusual for their background at all. The only way to know your actual list is evidence from your own speech, not a general prediction from your first language.

How the heatmap actually builds up

Each time you record a sentence, every phoneme in it gets scored individually and rolled into a running, recency-weighted picture across all 44 sounds — not just a single take's result, but a pattern across your practice history. A sound that fails once might be noise; a sound that fails repeatedly, across different words and different sessions, is a real pattern. That distinction — one flagged instance vs a repeated pattern — is exactly what a heatmap built over time can show that a single result can't.

Reading the map once you have one

Look for two things specifically: which cells are consistently red or flagged across multiple sessions (your real problem sounds), and whether flagged cells cluster around a recognizable category — several vowels in the same region of the mouth, or several consonants that share a place of articulation — since a cluster often points to one underlying articulation habit rather than several unrelated ones. A learner whose map shows both /θ/ and /ð/ flagged, for instance, likely has one underlying issue (the dental fricative tongue position) rather than two separate problems.

Sounds that show up cleanly across every session are worth leaving alone — practicing an already-solid sound doesn't move the needle, and time spent there is time not spent on the sounds that actually need it.

What to do with your list

Take the top two or three most consistently flagged sounds and work through them one at a time using the focused approach in why practicing one sound for a week works — don't try to fix all of them simultaneously, since that spreads attention thin across sounds that each need concentrated, spaced repetition to actually consolidate.

Drill it

There's no single drill sentence here by design — the whole point of this approach is that your list is personal. Start by recording two or three varied sentences you haven't specifically rehearsed, covering a range of sounds, and see what the results actually flag before choosing your first target.

How do I know it's working?

Watch for the total number of consistently-flagged sounds shrinking over successive weeks, not for the map turning uniformly green overnight — a map with two remaining red cells after starting with five is real, measurable progress, even though it's not yet a "finished" chart.

Try it

Open sayit free — no install, no card — and record a few varied sentences to start building your own picture. The features page covers how the per-phoneme history and heatmap fit into the rest of the product if you want the full picture.