1 · Tones 声调
English uses pitch for emotion; Mandarin uses it for meaning. Learn the four tones (+ neutral) before anything else — with your hand and your voice, not the alphabet.
Five degrees of pitch · 1st = 55, 2nd = 35, 3rd = 214, 4th = 51. In connected speech a 3rd tone before another 3rd tone is usually a half third (21): it dips and does not rise back up — trace the full 214 only when saying the syllable alone.
The m-scale — sing it, exaggerate it:
Minimal pairs — where tone changes the word
Click a word to hear it. Then run the quiz below: listen and pick the word you heard.
Tone 3 vs 4 — the classic shopping mix-up
The m-scale celebrity pair
Tone 3 vs 4
Tone 2 vs 4
Tone 1 vs 4
Tone 3 vs 4
Tone 4 vs 3 — asking vs kissing!
Tone 2 vs 4
Tone 1 vs 4
Tone 1 vs 4
Tone 1 vs 3
Tone 1 vs 4
Aspiration · b/p — same final and tone, only the puff of air changes
Aspiration · g/k — unaspirated vs aspirated velar stop
Retroflex vs flat · zh/z (the following i is /ʅ/ vs /ɿ/, so this pair also trains the apical vowel)
Nasal coda · -n (tongue tip) vs -ng (tongue root)
Rounding · ü vs u — say “ee” with rounded lips
Palatal series · j/q differ only in aspiration
🎧 Tone quiz
🎯 Tone shooter
2 · Initials 声母
Pinyin reuses English letters but not English sounds. The biggest trap: b / d / g are unaspirated (like “spy / stay / sky”), while p / t / k carry the strong puff of air.
b — both lips
Press both lips together, then release. Your tongue stays low and relaxed — the obstruction is at the lips, not the tongue.
💨 Air flows out of the mouth — no puff (unaspirated).
👆 Click any initial card above to see where the tongue goes.
Paper test: hold a tissue in front of your mouth — p / t / k / q / ch / c make it flap, b / d / g / j / zh / z don’t. That puff is aspiration, and it’s the only difference between the pairs.
3 · Finals 韵母
Compound finals are glides, not two separate vowels — your mouth slides from one shape to the other without stopping.
Click a compound final below
The chart shows the vowel space of your mouth — front/back is tongue position, high/low is jaw openness. A glide is one continuous slide along a path like this.
Simple finals
Compound finals · glide!
Nasal finals · front vs back
Front vs back nasal: -n endings (an / en / in) — tongue tip touches the upper gums, like the end of English “sun”. -ng endings (ang / eng / ing / ong) — tongue root presses the soft palate, like “song”. Hum both and feel where the buzz sits.
4 · Spelling makeover 拼写规则
Pinyin spells some syllables against intuition. The five classic makeovers below explain the odd letter swaps; the four rules after them are the orthography rules (apostrophe, tone-mark position, original tone, word spacing) that textbooks often skip.
i → y
i + vowel: add or swap in y
ian → yan烟 yān smokeiang → yang羊 yáng sheepi → yi一 yī one
u → w
u + vowel: add or swap in w
ua → wa蛙 wā froguo → wo我 wǒ meu → wu屋 wū house
ü → yu
ü always wears a yu disguise on its own
ü → yu鱼 yú fishüe → yue月 yuè moonüan → yuan元 yuán dollar
iu / ui / un
squeezed spellings of iou / uei / uen — the middle sound hides
iou → iu六 liù sixuei → ui水 shuǐ wateruen → un问 wèn ask
j q x + ü
after j, q, x the dots vanish: ju = jü — ü is never written with dots here
jü → ju局 jú gameqüe → que缺 quē lackxü → xu须 xū must
a / o / e → '
apostrophe before a later syllable starting with a, o or e — basis: 汉语拼音方案(1958) 隔音符号段; GB/T 16159 has no §4.6; §6.6.2 only covers line breaks
Xīān → Xī'ān西安 Xī'ān — without the apostrophe it reads as 希安kùài → kù'ài酷爱 kù'ài to love deeplynǚér → nǚ'ér女儿 nǚ'ér daughter
Tone mark position
a > o > e > i > u > ü — except iu / ui, which mark the LAST vowel (GB/T 16159 §6.5.1)
hao3 → hǎoa wins: 好 hǎoliu4 → liùiu → mark u: 六 liùhui4 → huìui → mark i: 会 huì
Write the ORIGINAL tone
spelling keeps the dictionary tone; connected speech changes it (GB/T 16159 §6.5.2)
yī gè → yí gè (spoken)一 before a 4th tone is read yí but still written yīnǐ hǎo → ní hǎo (spoken)two 3rd tones: the first rises, the spelling stays nǐma / de / le → no tone markneutral-tone particles are written unmarked
Write WORDS, not characters
join syllables of one word; split surname from given name; capitalize proper nouns
pīn yīn → pīnyīnone word → one block (§5.1)Běi jīng → Běijīngplace names stay joined (§6.2.2.3, capital per §6.3.2)Lǔ Xùn → Lǔ Xùnsurname | given name (§6.2.1.1)
5 · Whole syllables 整体认读
These 16 blocks don’t split. In zhi / chi / shi / ri the i is a retroflex apical vowel /ʅ/; in zi / ci / si it is a flat-tongue apical vowel /ɿ/. They are different sounds — and neither is English “ee” nor the vowel of “sure”. Each card gives the plain shape plus a toned demo: the character you see is pronounced exactly as written, and the recording uses that same reading.
Erhua (儿化) — bonus for later: Beijing speech adds an r-color to endings (花儿 huār “flower”). You don’t need to produce it at first — just recognize it when you hear it.
Syllable chart 音节表
A syllable grid: initials × finals. Which cells exist is decided by the standard Mandarin initial–final combination table (about 400 toneless syllables × 4 tones), not by whether we can find an example character. Pick a tone, then click a cell to hear it. A cell marked “—” means the syllable is legal, but our dictionary has no unambiguous common character at that pitch — most of these are genuine tonal gaps.
Example characters come from CC-CEDICT filtered by word frequency; the character shown and the character used for the recording share exactly the same initial, final and tone (verified at build time). A blank example never means the syllable is invalid. Cells with a studio clip play a recorded voice; cells marked “·TTS” fall back to your browser’s zh-CN speech voice, driven by that same character (never by Latin letters, which a Chinese voice would read as English). Tonal gaps are silent by design. Spelling follows 《汉语拼音方案》(1958) and GB/T 16159-2012.
Coverage, in counts (not percentages): GF0025-2021 lists 1110 syllables — 1094 with a tone mark and 16 neutral-tone. This grid has 1632 cells (408 toneless syllables × 4 tones); 1094 official syllables map to a cell and 538 grid cells are outside the official list. The two counts use different conventions — a cell outside the list is not an error, and this filter is a study aid, not a proficiency claim.
Drill 实战
Build any syllable from its parts, then face the ten traps English speakers fall into most.
1 · Syllable builder 音节拼合器
Pick an initial, a final and a tone. If the syllable exists in Mandarin you’ll see it with its most common character — and hear it.
2 · Trap challenge 错误词库挑战
Where English speakers slip most: retroflex vs flat, ü vs u, aspiration and tone sandhi. Every miss explains itself.
Questions are drawn from the selected band and weighted toward the features you miss most. Items with a spoken syllable follow GF0025-2021; rule questions are placed by this site’s teaching order (not an official grade). The report lists which distinctive features you distinguished and which to revisit — it is not a CEFR level or a proficiency certificate.
Basis & scope 依据与范围
What the spelling on this site is based on — and what it does not claim.
This is a pronunciation foundation course, not graded by CEFR or HSK. CEFR phonological-control scales describe overall intelligibility, not which initial or syllable belongs to A1 or B2, so no sound or syllable here carries a CEFR label. The Study band filter uses the official GF0025-2021 graded syllable list (1110 syllables; elementary 608 / intermediate 300 / advanced 202) only to suggest a learning order. It is a study aid: it does not place you at a level, does not convert to CEFR, and is not a certificate. Rule questions without a spoken syllable are ordered by this site’s teaching sequence, not by an official grade.
Syllable spelling and tone marks follow the 《汉语拼音方案》(1958). Word-level joining, separators, capitalisation and tone marking follow GB/T 16159-2012《汉语拼音正词法基本规则》; personal names additionally follow GB/T 28039-2011. ISO 7098:2015 is consulted for reference only. These GB/T standards are recommendatory national standards: there is no compliance certificate, and this site does not claim one.
Written pinyin on reading pages keeps original tones (e.g. yī ge), as GB/T 16159-2012 §6.5.2 allows; the spoken sandhi (yí ge) is taught in the Trap challenge above. CEFR labels used elsewhere on this site are an approximate guide to HSK, not an official conversion. Some short phrases and rare characters still fall back to per-character annotation — see the site's annotation spec for the full list of known deviations.
Pronunciation demos: tone contours use the standard five-degree values 55 / 35 / 214 / 51; articulatory descriptions follow modern Mandarin phonetics (e.g. zh is a voiceless unaspirated retroflex affricate, and the i after zh/ch/sh/r/z/c/s is an apical vowel, not English “ee”). Every demo character’s dictionary reading is checked against its label at build time, and recordings are only served when their text matches that character. These are checks on the text we send to the voice engine; the engine’s own audio quality is outside this site’s control.