Chinese AI Voice

HsiaoChen — free Taiwanese Chinese text to speech

HsiaoChen is a free female Taiwanese Chinese AI voice — a Microsoft neural voice. Type your text below to hear it read aloud, then download the audio. No account needed.

Female Taiwanese accent Global voice Free

Make HsiaoChen say your words

Voice details

Language
Chinese
Accent
Taiwanese
Gender
Female
Type
Global · Microsoft neural
Download
MP3
Speed
Natural pace
Voice ID
zh-TW-HsiaoChenNeural
Price
Free

How HsiaoChen sounds — measured

Median pitch

242 Hz

In the middle of 8 female Chinese voices

Intonation range

10.3 semitones

Close to typical for its group

Tone (spectral centroid)

2,412 Hz

Balanced tone for its group

Speaking rate

4.7 characters/s

In the middle of 14 Chinese voices on the same passage

Pitch across the sample passage dashed line = HsiaoChen’s median

How long your audio will be at 4.7 characters/s

1,000 characters
≈ 3 min 34 s
5,000 characters (free limit per render)
≈ 18 min
  • ↔Compared with HsiaoYu, HsiaoChen has a higher pitch (242 vs 202 Hz) and faster delivery (4.7 vs 3.9 characters/s). Listen to HsiaoYu →
  • ↔Compared with Xiaoxiao, HsiaoChen has a similar pitch (242 vs 243 Hz). Listen to Xiaoxiao →

Measured by SpeakScribe, October 2026 · How we measure · Every Chinese voice side by side

Use HsiaoChen in a conversation

With Multi-voice, HsiaoChen can read one character while Yunxi (a male Chinese voice) reads another — the whole dialogue comes back as one MP3 with even volume between speakers. A free account can make 3 two-speaker dialogues a month; paid plans allow up to 4 speakers on Starter.

Try a dialogue with HsiaoChen →

Formats, limits and help

HsiaoChen: quick answers

Measured from its audio, HsiaoChen has a median pitch 242 Hz (in the middle of 8 female Chinese voices); intonation range 10.3 semitones (close to typical for its group); tone (spectral centroid) 2,412 Hz (balanced tone for its group).
About 3 min 34 s. HsiaoChen reads about 4.7 characters a second, spaces included (measured on its sample passage), so the free limit of 5,000 characters per render is about 18 minutes of audio. Numbers, names and long pauses make it longer.