Chinese Text to Speech Is Harder Than It Looks. I Tested 5 Tools to Prove It.

There's a reason Chinese text to speech took longer to get right than English. Mandarin has four tones plus a neutral tone, and getting the tone wrong doesn't just sound bad — it changes the meaning of the word entirely. 买 (mǎi, third tone, to buy) and 卖 (mài, fourth tone, to sell) are opposite meanings separated by a tonal inflection that most TTS engines from five years ago would botch half the time. Then there are polyphones — characters that change pronunciation based on context. 行 is "xíng" when it means "okay" and "háng" when it means "row" or "bank." A good Chinese TTS engine needs to read the sentence, understand the context, then pick the right pronunciation. A bad one just guesses.

I spent two weeks running Chinese text through every free TTS tool I could find. I used the same test material for all of them — a 2,500-character Zhihu answer about Chinese education, a WeRead chapter from 三体 (The Three-Body Problem), and a Feishu document a colleague shared about project planning. The results were not close.

CastReader is the tool I didn't expect to care about and now can't stop using. It's a Chrome extension designed to read web pages aloud, and on most Chinese sites it does exactly what you'd want — click the icon on a Zhihu page, it extracts the answer text, skips the sidebar recommendations and comment noise, and reads with Kokoro AI voice while highlighting paragraphs on the actual page. The voice handles tones well. Polyphones are mostly correct. But the thing that makes CastReader genuinely unique is WeRead support. 微信读书 renders all its text as Canvas — the words you see on screen aren't HTML text, they're pixels painted onto a canvas element. Every other TTS tool I tried on WeRead saw a blank page. Nothing to read. CastReader intercepts the chapter data before it hits the Canvas renderer and reads the actual book text. I listened to two full chapters of 三体 on WeRead through CastReader and it worked. No other Chrome extension does this. Not one. It also handles Feishu documents, Yuque pages, and WeChat articles (mp.weixin.qq.com) with the same paragraph highlighting. No signup, no limits, free.

Google Translate's speaker icon is what most people reach for first. Paste Chinese text, click the icon, listen. The voice quality is good enough for short passages — sentence-level pronunciation checks, individual word lookups, a paragraph you want to hear once. Google's neural model handles most tones correctly and gets common polyphones right. The ceiling is 5,000 characters per paste, which is about 2,500 words in Chinese — enough for a short article but not a book chapter. And the workflow for full articles is painful. Open article. Select text. Copy. Switch to Google Translate tab. Paste. Click speaker. Go back. Select next section. Copy. Switch. Paste. Click. By the third round trip you'll understand why I went looking for alternatives.

Edge Read Aloud handles Chinese through Microsoft's Azure neural voice Xiaoxiao, which is one of the better Chinese TTS voices available anywhere. Open a Chinese web page in Edge, press Ctrl+Shift+U. The voice is natural, the tones are accurate, the pacing feels right for Mandarin prose. Word-level highlighting follows along. I read a full People's Daily editorial through Edge and it sounded like a professional news anchor. The problem — and it's the same problem Edge Read Aloud has for every language — is that you need Edge. If your browser is Chrome, if your bookmarks and extensions and passwords live in Chrome, switching to Edge for TTS means maintaining two browsers. Some people are fine with that. I'm not.

macOS Speak Selection with a downloaded Chinese Siri voice is the option nobody knows about. System Settings, Accessibility, Spoken Content, Speak Selection. Download Ting-Ting or one of the other Chinese voices. Now highlight Chinese text anywhere on your Mac — Safari, Preview, Notes, WeChat Desktop — press Option-Escape, and it reads aloud. The voice quality is a step behind the AI models but completely usable. On iPhone, the same Chinese voices work through Speak Selection in accessibility settings. My mother uses Speak Selection on her iPad to listen to WeChat articles friends send her. She doesn't know what TTS stands for and doesn't care. She highlights, taps Speak, listens while cooking. That's the target user experience.

Android handles Chinese TTS through Google's built-in engine. Enable Select to Speak in Accessibility settings, tap the floating button on screen, drag over Chinese text. The voice quality varies by device — flagship Pixels and Samsung Galaxy phones with updated Google TTS data sound good, budget phones from 2022 sound notably worse. If you're on Android and want Chinese TTS that works everywhere, check that your Google TTS is updated in the Play Store first. The difference between outdated and current voice data is the difference between tolerable and actually pleasant.

For reading Chinese web content — Zhihu answers, WeChat articles, online novels, Feishu docs — CastReader in Chrome is where I'd start. It's the only option that handles WeRead's Canvas rendering, and the paragraph highlighting makes following along in Chinese characters significantly easier than audio alone. For quick pronunciation checks, Google Translate is always a tab away. For the best raw voice quality on desktop, Edge Read Aloud with Xiaoxiao is hard to beat if you don't mind using Edge. All free. All actually capable of reading real Chinese text without mangling the tones.