Chinese Text to Speech

14 free Chinese AI voices

SpeakScribe has 14 Chinese voices in three groups: mainland Mandarin, Taiwanese Mandarin and Hong Kong Cantonese. Choose the region first.

Xiaoxiao

Female · Mainland China

GLOBAL
Use voice

Xiaoyi

Female · Mainland China

GLOBAL
Use voice

Yunxi

Male · Mainland China

GLOBAL
Use voice

Yunjian

Male · Mainland China

GLOBAL
Use voice

Yunxia

Male · Mainland China

GLOBAL
Use voice

Yunyang

Male · Mainland China

GLOBAL
Use voice

Xiaobei (Liaoning)

Female · Mainland China

GLOBAL
Use voice

Xiaoni (Shaanxi)

Female · Mainland China

GLOBAL
Use voice

HsiaoChen

Female · Taiwanese

GLOBAL
Use voice

HsiaoYu

Female · Taiwanese

GLOBAL
Use voice

YunJhe

Male · Taiwanese

GLOBAL
Use voice

HiuGaai

Female · Hong Kong

GLOBAL
Use voice

HiuMaan

Female · Hong Kong

GLOBAL
Use voice

WanLung

Male · Hong Kong

GLOBAL
Use voice

How Chinese is written, and what to type so the voice reads it right

Chinese has no spaces between words, so the voice works out word boundaries from the characters. Type the script as a native writer would, with full-width punctuation: , and 。 give the clearest pauses. Match the character set to the voice: simplified characters (简体) for the mainland voices, traditional characters (繁體) for the Taiwan and Hong Kong voices.

Many characters have more than one reading, and the voice chooses from context. When one is misread, replace the word with another that has the same meaning rather than adding pinyin.

  • Simplified for mainland voices: 这个声音. Traditional for Taiwan and Hong Kong voices: 這個聲音.
  • Polyphonic characters: 行 is háng in 银行 (bank) and xíng in 行走 (to walk); 长 is zhǎng in 长大 (to grow up) and cháng in 长度 (length); 重 is chóng in 重新 (again) and zhòng in 重要 (important).
  • Type characters, not pinyin. Tone changes such as 不 before a fourth tone (不是 búshì) are part of normal reading, so no tone marks are needed.

Numbers, dates and loanwords in Chinese text

Digits are common in Chinese text, but some readings depend on the counter that follows. Write 两个 rather than 2个 when the word for two must be read as liǎng. Years are conventionally read digit by digit, so write 二零二六年 when you want that reading exactly.

  • Write dates as 2026年10月22日 and times as 下午三点半 (3:30 p.m.).
  • Spell out a percentage where the number matters: 百分之五十 instead of 50%.

Accents and regional varieties

The voices here cover three: Mandarin as spoken in mainland China, Mandarin as spoken in Taiwan, and Cantonese as spoken in Hong Kong. Cantonese is a separate spoken language with its own tones, so write each script in the variety its voice speaks.

Standard written Chinese read in Cantonese sounds formal, like a news report. For a conversational Hong Kong style, write colloquial Cantonese with characters such as 嘅, 咗, 喺 and 啲. Taiwanese Mandarin uses little of the northern 儿 (erhua) ending, so leave 儿 out of scripts for Taiwan.

Choosing between the Chinese voices

The eight mainland Studio voices (Xiaobei, Xiaoni, Xiaoxiao, Xiaoyi, Yunjian, Yunxi, Yunxia and Yunyang) were retired in October 2026. In a speech-recognition check, 45–82% of the characters they read came back wrong, against 0–2% for the Global voices, so their old pages now lead to the Global voice of the same name.

Mainland Mandarin has eight Global voices: Xiaoxiao, Xiaoyi, Yunxi, Yunjian, Yunxia and Yunyang, plus two regional voices. Global Yunxia (male, 336.3 Hz) is the highest-pitched voice in the set. Global Yunyang (male, 106.5 Hz, 13.8 semitones) is the lowest, and Global Yunjian (male, 122.3 Hz, 16.8 semitones) has the largest range in the set. The regional voices Xiaobei (Liaoning) (female, 208.8 Hz) and Xiaoni (Shaanxi) (female, 218.4 Hz) carry the regional flavour named in their titles.

For Taiwan, HsiaoChen (female, 242.1 Hz, 10.3 semitones) and HsiaoYu (female, 202.2 Hz, 9.1 semitones) are the female voices, and YunJhe (male, 117.5 Hz, 11.2 semitones) is the male voice. For Hong Kong Cantonese, HiuGaai (female, 210.9 Hz, 5.8 semitones) and HiuMaan (female, 264.6 Hz, 7.1 semitones), and WanLung (male, 112.7 Hz, 7.5 semitones) is male.

Common mistakes when turning Chinese text into speech

Many Chinese script problems come from the character set and the punctuation. The fixes below prevent most misreadings.

  • Mixing simplified and traditional characters in one script, such as 这 and 這 in the same paragraph. Pick one set for the whole script, matching the region of the voice.
  • Pasting pinyin with tone numbers, such as ni3 hao3, instead of 你好. Latin letters are read as Latin letters, so type the characters.
  • Using half-width punctuation such as , and . in Chinese sentences. Full-width , and 。 mark the pauses that Chinese readers expect.
  • Adding spaces between Chinese words. Chinese is normally written without them, and extra spaces create pauses that the sentence does not need.

Chinese voices compared — measured

Each voice read the same Chinese text at normal speed. Lower pitch sounds deeper; a wider intonation range sounds livelier, a narrower one steadier.

VoiceGenderAccent PitchIntonation range Speaking rate
HsiaoYu GLOBAL Female Taiwanese 202 Hz 9.1 st 3.9 chars/s
Xiaobei (Liaoning) GLOBAL Female Mainland China 209 Hz 10.7 st 4.2 chars/s
HiuGaai GLOBAL Female Hong Kong 211 Hz 5.8 st 3.4 chars/s
Xiaoni (Shaanxi) GLOBAL Female Mainland China 218 Hz 12.6 st 4.8 chars/s
HsiaoChen GLOBAL Female Taiwanese 242 Hz 10.3 st 4.7 chars/s
Xiaoxiao GLOBAL Female Mainland China 243 Hz 9.0 st 4.9 chars/s
HiuMaan GLOBAL Female Hong Kong 265 Hz 7.1 st 4.6 chars/s
Xiaoyi GLOBAL Female Mainland China 306 Hz 7.5 st 5.0 chars/s
Yunyang GLOBAL Male Mainland China 106 Hz 13.8 st 5.4 chars/s
WanLung GLOBAL Male Hong Kong 113 Hz 7.5 st 4.8 chars/s
YunJhe GLOBAL Male Taiwanese 118 Hz 11.2 st 5.0 chars/s
Yunjian GLOBAL Male Mainland China 122 Hz 16.8 st 4.8 chars/s
Yunxi GLOBAL Male Mainland China 142 Hz 10.4 st 5.0 chars/s
Yunxia GLOBAL Male Mainland China 336 Hz 10.2 st 4.7 chars/s

Measured by SpeakScribe in October 2026 from each voice's audio. Pitch is the median fundamental frequency; intonation range is the spread between the 10th and 90th percentile of pitch. Speaking rate is characters per second, spaces included, on the ~15-second sample passage that every Chinese voice reads. st = semitones. Methodology · How SpeakScribe voices work

More for Chinese creators

Frequently asked questions

Yes. HiuGaai and HiuMaan (female) and WanLung (male) are Hong Kong Global voices that speak Cantonese.
Yes: HsiaoChen and HsiaoYu (female) and YunJhe (male). Write the script in traditional characters.
14: eight mainland Global voices, three Taiwanese and three Hong Kong voices. The eight mainland Studio voices were retired in October 2026.
No. Pinyin without characters is read as Latin letters. Type the script in Chinese characters, and no tone marks are needed.
Yes. See Chinese audio to text.
No. The library has no voices for other Chinese varieties such as Shanghainese or Hokkien.
Yes. Keep English words in Latin letters with a space on each side, and preview the line before you generate.

Generate Chinese voiceover audio free

Pick a voice above, paste your script, and export publish-ready audio in seconds.

Open Voice Studio