Hayane Pinyin

Hayane Pinyin trains the most overlooked yet pedagogically critical contrast in Mandarin pronunciation: aspiration.
The pinyin pairs b/p, d/t, g/k look familiar to learners worldwide, but they don't work the way you think. Both members of each pair are voiceless. The difference is whether breath escapes after the burst — what phoneticians call Voice Onset Time (VOT).
Most learners produce these stops by mapping them to similar-looking sounds in their first language (Japaneseば for pinyin b, English p for pinyin p). The result is systematic, hard-to-correct accent errors that ear-based feedback alone won't catch.
Hayane Pinyin solves this with three things textbooks
cannot:
• Real-time VOT measurement — your pronunciation gets an
exact millisecond reading the moment you finish speaking.
• Three-way verdict (good unaspirated / borderline / good aspirated), per phoneme, with research-grounded thresholds.
• Visual feedback: a VOT bar showing where you landed relative to native targets, plus the actual waveform of your recording with burst and voicing-onset markers.
Special detection for the most common Japanese-learner error: voiced ば substitution (pre-voicing, negative VOT) is flagged in red so you can correct it directly rather than guess.
Each example word is read out by high-quality on-device
Mandarin speech synthesis, so you can hear the model and immediately mimic. How it works:
1. Choose a phoneme pair (b/p, d/t, g/k).
2. Listen to the model word.
3. Tap record and produce the syllable.
4. Read your VOT value, the three-way verdict, and the burst-to-voicing waveform.

UI available in English, Japanese, Korean, Simplified Chinese, and Traditional Chinese.

Privacy first
• Microphone audio never leaves your device — all analysis is on-device.
• No accounts, no analytics, no third-party SDKs, no tracking.
• Works fully offline.
Built on the academic phonetics literature (Chen, Chao & Peng 2007; Li 2013; Lisker & Abramson 1964; Ladefoged & Maddieson 1996) as the design foundation, with threshold derived from native Mandarin distributions.