Does sayit work well for a Japanese accent?

sayit works for a Japanese accent through what its engine is specifically built to catch: inserted vowels around consonant clusters. Japanese syllable structure doesn't allow most consonants to stand next to each other, so English clusters get an extra vowel wedged in ("desk" tending toward "de-su-ku") — sayit's scoring is built to name this class of error directly rather than let a lenient transcript absorb it, because it's measuring the phonemes you produced, not just whether a listener could still guess the word.

30-second version: Japanese isn't one of the four accent groups sayit names as explicitly tested (Indian, American, British, European), so treat this as a mechanism-based answer, not a published benchmark. The engine is designed to name inserted-vowel patterns directly, and its phoneme-vs-target-IPA scoring is a better fit than a word-recognition tool for the /r/–/l/ contrast that's famously difficult when your first language uses a single flap sound for both.

Why the mechanism is the honest answer here

The two hardest things about Japanese-influenced English pronunciation — inserted vowels breaking up consonant clusters, and the /r/–/l/ distinction collapsing into one sound — both share a property: a speech-recognition model can often still guess the intended word even when either happens, because context does the heavy lifting. That's precisely why a transcript-based tool tends to score these as fine. sayit measures the phonemes against the target directly, so an inserted vowel or a merged /r/-/l/ shows up as a real, named gap rather than getting smoothed over by a model whose job is to be forgiving.

Where this is genuinely strong

• Per-word, per-phoneme feedback with target IPA next to your actual sound makes an inserted vowel or an /r/–/l/ substitution visible instead of silently accepted.
• Every core drill (pacing, vowel contrasts, minimal pairs) plus read-aloud scoring is free, so you can test the /r/–/l/ contrast specifically with no card.
• IELTS and TOEFL exam-format practice (Pro/Max) is available if you're preparing for either exam.

Where to keep expectations honest

• Japanese isn't among the four accent groups sayit publicly names as tested — this is an argument from mechanism, not a published Japanese-specific benchmark.
• Pitch accent, a defining feature of Japanese that differs from English stress, isn't a separately named analysis — sayit's feedback covers phoneme clarity and general intonation, not a dedicated pitch-accent transfer check.
• It's pronunciation and fluency only — no grammar or vocabulary scoring, which is a separate source of Japanese-English interference (articles, plurals).

How to test the fit on your own

Record a short list of words with English consonant clusters at the start or end ("desk," "student," "strong") alongside a couple of /r/–/l/ minimal pairs, and look at whether the per-word feedback names the inserted vowel or the merged consonant specifically, rather than just returning a low overall score with no explanation. That specificity — a named sound, not just a number — is the actual test of whether a scoring tool is useful for this particular pattern, and it costs nothing to check on the free plan.

Who this is for

If inserted vowels or the /r/–/l/ merge are the specific things you're trying to work on, a minimal-pair drill (ship versus sheep, rice versus lice) paired with per-phoneme scoring is directly useful, and that's a core sayit feature, free on every tier.

Try it

Record a word with a consonant cluster or an /r/–/l/ pair in the app — no card needed for the first take — and see exactly what it flags.