What happens when I tap a flagged word in sayit?

Tapping any word in your results opens it into a per-word inspector — a breakdown of the word into its individual sounds, each one color-coded by whether it matched, and, for anything flagged, the target phoneme next to the one sayit actually detected. It's the difference between being told a word scored 71% and being shown exactly which of its five or six sounds is the reason.

30-second version: Any word in your transcript is a tap target. Opening it expands the word into its phoneme atoms, colored by match, substitution, insertion, or deletion, so you can see at a glance which specific sound (not just which word) needs work. Flagged phonemes show their target and produced IPA together, and you can hear the isolated sound rather than only the whole word.

Why inspect a word instead of just reading its score?

Because a word-level score is already a summary, and a five-phoneme word with one wrong sound and a five-phoneme word where every sound is slightly off can land at a similar score while needing completely different fixes. The inspector removes that ambiguity: instead of inferring where the problem probably is, you see the word broken into its actual sound units, with the ones that didn't match visibly marked apart from the ones that did.

What does the color-coding inside a word actually mean?

Each phoneme atom in the word — not the word as a whole — gets its own status. A match stays neutral; a substitution (you produced a different sound than expected) and a deletion (a sound was dropped) are marked distinctly, since they're different problems with different fixes. This is the same underlying alignment used to build the target-versus-produced IPA comparison, just laid out sound by sound rather than as two full strings.

Can I hear the individual sound, not just the whole word?

Yes — isolated phones are audible on their own, not only as part of the whole word. Hearing a single consonant or vowel in isolation is often more useful for correcting it than hearing it inside a fast word, because the surrounding sounds in a word can mask exactly what's different about the one that's wrong.

Does the inspector tell me what to do differently, or just what was wrong?

Both. Alongside the phoneme-level breakdown, a flagged sound typically carries a short articulation cue — what to do with your tongue, lips, or voicing to move that specific sound closer to the target — the same guidance behind the one-fix recommendation, but scoped to the single word you tapped rather than the whole take.

How is this different from just seeing the whole take's transcript?

The full transcript view is built for scanning — it shows every word's verdict at a glance so you can see the shape of a whole passage at once. The inspector is built for one word at a time, in depth: it's what you open once the transcript view has already told you which word deserves a closer look. Between them, they cover both "what's the overall picture" and "what exactly happened here."

What you see at each level

Try it

Read a sentence in sayit and tap any word marked unclear or check in your results — the word opens into its own sound-by-sound view. If you want to re-record just that word after seeing what went wrong, that's covered in rechecking a single word.