Pronunciation Lab

🇬🇧 Step-by-Step: Diagnose & Drill Your Weak Phonemes

① Open "Pronunciation Lab" from the Speak menu. It hosts four trainers: Phoneme Diagnostic, Minimal Pairs Drill, Word Stress, and Intonation. ② Start with the Diagnostic (do this once). You record 15 short calibration sentences that together exercise all 44 English phonemes. Whisper transcribes each, then GPT scores every phoneme on a 0.0–1.0 accuracy scale. ③ The Diagnostic writes your PhonemeProfile: phonemes with currentAccuracy < 0.6 are flagged isWeak=true and become the priority list for every other Lab trainer. ④ Minimal Pairs Drill — auto-prioritises drills for your weakest contrasts (e.g. /ʃ/ vs /s/, /iː/ vs /ɪ/, /θ/ vs /ð/). For each pair you record one specific word; the scorer tells you which word it actually heard plus a confidence score. ⑤ Word Stress trainer — record multi-syllable words and get pass/fail on whether the correct syllable was stressed. ⑥ Intonation trainer — record short sentences and get pass/fail on whether your pitch contour matched the model. ⑦ Every drill updates your PhonemeProfile, so over time the Lab reshuffles to drill what still needs work — your weakest phonemes always rise to the top. ⑧ "Re-calibrate" anytime to refresh your baseline (e.g. after weeks of practice). ⑨ Tip: the Lab pairs perfectly with the new 2026 task types. Read Aloud, Repeat Sentence and TOEFL Listen & Repeat all reward intelligible phoneme production — Lab drills directly target the same skill, so a weekly Lab session lifts your scores across all three.