Skip to content

Tone guide

Mandarin, Cantonese, Vietnamese and Thai are the tonal languages with a model. Tone is a separate layer: the vowel /a/ with tone 35 and with tone 21 is the same phoneme token, and tone never moves a boundary. It sits on the syllable's vowel (on the last token when the vowel is split in two), and consonants carry none.

The tone comes from the G2P's dictionary, not from the audio: it is the word's lexical tone, with tone sandhi not applied, so the speaker may say something different. Change it only when you are sure.

Reading the numbers

Tones are written as Chao numbers: pitch on a scale from 1 (lowest) to 5 (highest) in the speaker's range. Two digits are where the pitch starts and ends, three digits add the turning point. One digit is a level pitch, used for short (checked) syllables and for level tones written compactly. The same number can be a different tone in another language: read the table of the language you are annotating.

Mandarin, Cantonese, Vietnamese, Thai

Mandarin (zh)

Four tones and a neutral tone. The G2P gives each character its dictionary reading, so tone sandhi is not applied (a 3rd tone before another 3rd tone is still 214, though it is said as 35) and particles may carry their full tone.

number tone pitch notes share examples
55 ˥˥ Tone 1, high level high level ā; yīnpíng 阴平 19.7% 一 /i55/; 它 /tʰ u ɔ55/; 中 /ʈ ʂ ʊ55 ŋ/
35 ˧˥ Tone 2, rising rising, mid to high á; yángpíng 阳平 26.7% 的 /t i35/; 和 /x ɤ35/; 人 /ʐ ə35 n/
214 ˨˩˦ Tone 3, dipping dipping: falls low, then rises ǎ; shǎngshēng 上声. Before another syllable it is often only low (21). 17.7% 了 /l j ɑ ʊ214/; 可以 /kʰ ɤ214 i214/; 有 /j o ʊ214/
51 ˥˩ Tone 4, falling falling, high to low à; qùshēng 去声 34.8% 在 /t s a ɪ51/; 是 /ʂ ɚ51/; 不 /p u51/
3 ˧ Neutral tone mid (one level) qīngshēng 轻声: short and unstressed; its pitch follows the tone before it 1.0% 一个 /i35 k ɤ3/; 他们 /tʰ a55 m ə3 n/; 男子 /n a35 n t s ɯ3/
5 ˥ (rare) high (one level) only in spelled-out letters and foreign words 0.0% npws /ɛ55 n p j ɑ ʊ55 ɛ5 s/

Cantonese (yue)

Six tones, numbered 1–6 in Jyutping. Level tones are written with one digit. The checked (entering) tones of syllables ending in p, t, k share the number of the level tone at the same height.

number tone pitch notes share examples
5 ˥ Tone 1, high level high (one level) Jyutping 1 (and the high entering tone, 7) 27.6% 的 /t ɪ5 k/; 都 /t o u5/; 方式 /f ɔ5 ŋ s ɪ5 k/
35 ˧˥ Tone 2, high rising rising, mid to high Jyutping 2 11.2% 影響 /j ɪ35 ŋ h œ35 ŋ/; 許多 /h ɵ y35 t ɔ5/; 使 /s i35/
3 ˧ Tone 3, mid level mid (one level) Jyutping 3 (and the mid entering tone, 8) 13.6% 世界 /s ɐ i3 k a i3/; 個 /k ɔ3/; 國家 /k w ɔ3 k k a5/
21 ˨˩ Tone 4, low falling low falling Jyutping 4 17.6% 成為 /s ɪ21 ŋ w ɐ i21/; 人 /j ɐ21 n/; 不同 /p ɐ5 t tʰ ʊ21 ŋ/
13 ˩˧ Tone 5, low rising low rising Jyutping 5 7.3% 有 /j ɐ u13/; 以 /j i13/; 這個 /t s ɛ13 k ɔ3/
2 ˨ Tone 6, low level low (one level) Jyutping 6 (and the low entering tone, 9) 22.7% 動物 /t ʊ2 ŋ m ɐ2 t/; 在 /t s ɔ i2/; 活動 /w u2 t t ʊ2 ŋ/

Vietnamese (Northern) (vi)

Six tones, shown by the accent marks of the spelling. Sắc is split in two: syllables ending in p, t, c, ch take a short high rise (45). Counts are from Northern Vietnamese (vi); vi-c and vi-s may number some tones differently.

number tone pitch notes share examples
33 ˧˧ ngang, level mid level no mark: a 29.0% trong /c ɔ33 ŋ͡m/; không /x o33 ŋ͡m/; cho /c ɔ33/
32 ˧˨ huyền, low falling mid falling grave: à 17.2% và /v a32/; là /l a32/; người /ŋ ɯ ə32 j/
24 ˨˦ sắc, rising rising, low-mid to high-mid acute: á 13.6% có /k ɔ24/; với /v ɤ24 j/; số /s o24/
45 ˦˥ sắc, high rising (checked) high rising acute on a syllable ending in p, t, c, ch: các, nhất 8.7% các /k a45 k/; nhất /ɲ ɤ̆45 t/; nước /n ɯ ə45 k/
21 ˨˩ nặng, low, glottalised low falling dot below: ạ 16.8% một /m o21 t/; được /d ɯ ə21 k/; động /d o21 ŋ͡m ʔ/
312 ˧˩˨ hỏi, dipping dipping: mid, low, then mid hook above: ả 10.2% của /k u ə312/; ở /ɤ312/; thể /tʰ e312/
35 ˧˥ ngã, broken rising rising, mid to high tilde: ã; often with a glottal break in the middle 4.3% những /ɲ ɯ35 ŋ/; đã /d a35/; sẽ /s ɛ35/
3 ˧ (rare) mid (one level) only in foreign words 0.1% martelly /m a r t ɛ3 l i33/; knp /k a33 n p3/; auckland /ă3 w k l a33 n d/

Thai (th)

Five tones. The tone of a syllable follows from its consonant class, vowel length, ending and tone mark; the G2P works it out.

number tone pitch notes share examples
3 ˧ mid tone mid (one level) siang saman 34.0% ใน /n a3 j/; มี /m i3/; การ /k a3 n/
21 ˨˩ low tone low falling siang ek 22.1% ต่าง /t a21 ŋ/; จะ /t͡ɕ a21 ʔ/; จาก /t͡ɕ a21 k̚/
51 ˥˩ falling tone falling, high to low siang tho 19.3% ที่ /tʰ i51/; ได้ /d a51 j/; ไม่ /m a51 j/
45 ˦˥ high tone high rising siang tri 15.2% และ /l ɛ45 ʔ/; นี้ /n i45/; ใช้ /t͡ɕʰ a45 j/
114 ˩˩˦ rising tone low, then rising siang chattawa: starts low, then rises 8.8% ของ /kʰ ɔ114 ŋ/; หรือ /r ɯ114/; สามารถ /s a114 m a51 t̚/
33 ˧˧ (rare) mid level only in foreign words 0.5% super-g / u p ɛ33 r ɛ33/; - / a33/; anna /a33 n a33/
35 ˧˥ (rare) rising, mid to high only in foreign words 0.1% jong /j ɔ35 /; robin / ɔ b i35 n/; range / a35 ŋ/

All 22 tone numbers

The tone head of every tonal model has these 22 classes (0 is “no tone”).

number letters pitch
35 ˧˥ rising, mid to high
21 ˨˩ low falling
33 ˧˧ mid level
45 ˦˥ high rising
51 ˥˩ falling, high to low
32 ˧˨ mid falling
214 ˨˩˦ dipping: falls low, then rises
55 ˥˥ high level
24 ˨˦ rising, low-mid to high-mid
3 ˧ mid (one level)
312 ˧˩˨ dipping: mid, low, then mid
5 ˥ high (one level)
13 ˩˧ low rising
2 ˨ low (one level)
212 ˨˩˨ low dipping
42 ˦˨ falling, high-mid to low
44 ˦˦ high-mid level
53 ˥˧ high falling
22 ˨˨ low level
4 ˦ high-mid (one level)
114 ˩˩˦ low, then rising

Shares and examples: FLEURS test sentences (dev split) through the same G2P (Mandarin 403 sentences, Cantonese 359, Vietnamese 361, Thai 439).