Mandarin vs Cantonese Text to Speech: Which to Choose
Pick Mandarin if your listeners are in mainland China or Taiwan, and Cantonese if they are in Hong Kong, Macau or Guangdong. Then write the script for that variety: standard written Chinese for Mandarin, and colloquial written Cantonese if you want Cantonese that sounds like everyday speech. Voxmint's Chinese text to speech page has a preview of every voice, so you can hear each variety before you write a word.
Start with the listener
Mandarin and Cantonese share a writing tradition but sound very different when spoken. A script written for one often reads strangely in the other, and a Hong Kong listener will notice immediately if a Cantonese video is read in Mandarin-flavoured speech. So the first question is who is listening:
- Mainland audiences, news, explainers, corporate content: Standard Mandarin. Jun, Haoran and Xiaoyu speak it.
- Taiwanese viewers, vlogs, lifestyle content: Taiwan Mandarin. Yating speaks it.
- Folk tales and children's stories with a regional feel: northern Mandarin. Guoqiang speaks it.
- Hong Kong and Cantonese-speaking audiences: Cantonese. Kaman is the Hong Kong Cantonese voice.
Mei, a soft and conversational voice, rounds out the seven. If your audience is mixed or you are not sure, Standard Mandarin is the safe default.
Standard written Chinese vs written Cantonese
This is where most Cantonese scripts go wrong. Formal written Chinese is built on Mandarin grammar and vocabulary. In Hong Kong it is read aloud in Cantonese pronunciation all the time, for news and official announcements, but nobody talks like that. Spoken Cantonese uses different words, different particles and different characters.
Here is the same idea written both ways:
| Standard written Chinese | Written Cantonese | Meaning |
|---|---|---|
| 他在哪裏？ | 佢喺邊度？ | Where is he? |
| 我們吃飯了沒有？ | 我哋食咗飯未？ | Have we eaten yet? |
| 這是我的。 | 呢個係我嘅。 | This one is mine. |
| 不要看。 | 唔好睇。 | Don't look. |
| 我現在沒有時間。 | 我而家冇時間。 | I don't have time now. |
| 你說什麼？ | 你講乜嘢？ | What did you say? |
The colloquial characters in the right-hand column (佢, 喺, 哋, 咗, 嘅, 係, 唔, 冇) are how people write the way they speak. Kaman's own sample uses this style: 今日天氣好好 instead of 今天天氣很好, 下晝 for "afternoon", and 帶定把遮 for "take an umbrella with you".
Which one should you feed the Cantonese voice?
It depends on the result you want.
- You want everyday Hong Kong speech (a vlog, an ad, a friendly radio-style read, dialogue): write in colloquial Cantonese. A script in formal Mandarin wording is the register of a news bulletin, so expect the read to sound like one, or test whether the voice keeps to Cantonese at all.
- You want a formal read (an announcement, a notice, something that would be read from a written text anyway): standard written Chinese is closer to what listeners expect from that kind of content. Generate a short test and listen before committing to it.
If you are translating from English or from a Mandarin script, ask the translator for a colloquial Cantonese version. A character-by-character conversion of Mandarin text gives you Mandarin wording with Cantonese sounds, which is the stiff result you are trying to avoid.
Traditional or simplified characters?
Hong Kong and Taiwan use traditional characters; the mainland uses simplified. The voice follows the characters you give it, and both Kaman's and Yating's samples were written in traditional characters. For Cantonese, traditional is what most Hong Kong readers use and is the natural choice. If your Cantonese script is in simplified characters, run a short test first.
Choosing between the Mandarin varieties
Even within Mandarin, the variety changes the feel.
- Standard Mandarin is neutral and clear. Use it when you need a broad mainland audience to hear the script without any regional flavour.
- Northern Mandarin carries the 儿化 (erhua) sound heard in Guoqiang's sample, as in 这事儿. It suits warm storytelling.
- Taiwan Mandarin sounds softer and more relaxed. Taiwanese viewers also use different everyday vocabulary, so write for them: 影片 rather than 视频 for "video", 軟體 rather than 软件 for "software", 計程車 rather than 出租车 for "taxi". Sentence-final particles like 喔 and greetings like 哈囉, both in Yating's sample, help the read sound local.
Script tips that apply to both
- Write numbers the way you want them spoken. A year and a quantity are read differently, so write 二零二六年 for a year rather than a bare 2026.
- Characters with more than one reading (行, 长, 重 and similar) are usually resolved from context. When one comes out wrong, rewording the phrase is normally faster than fighting it.
- Names and rare characters are the most likely place for a tone slip. If a name matters, listen to the take before publishing and add a short phrase around it, or re-generate.
- Punctuation controls pacing. A comma gives a short breath, a full stop a proper pause, and an exclamation mark lifts the energy. Split long sentences.
- English words. Short English words and brand names inside a Mandarin sentence usually read naturally. Hong Kong speech mixes in English words often, so if your Cantonese script does too, test a few lines with Kaman first.
When the catalogue doesn't have the voice you need
There is one ready-made Cantonese voice. If you need a different one, such as a male Cantonese narrator, describe it in words and let Voxmint design the voice, naming the variety in the description. Generate a few takes and keep the one you like. For a long series where every episode must sound like the same speaker, a ready-made catalogue voice is the most consistent choice.
FAQ
Is Cantonese just Mandarin with a different accent?
No. They are different spoken varieties of Chinese with different sounds, vocabulary and grammar. A Mandarin speaker and a Cantonese speaker who know only their own variety will struggle to understand each other in speech.
Can I give the Cantonese voice a Mandarin script?
You can try, but standard written Chinese is the register of news bulletins, so at best it will sound formal. For natural everyday Cantonese, write the script in colloquial Cantonese characters. Generate a short test and listen before you commit to either.
Should I use Taiwan Mandarin or Standard Mandarin for Taiwanese viewers?
Taiwan Mandarin. Yating speaks it, and pairing her with traditional characters and Taiwanese vocabulary gives the most local result.
Which variety should I use if my audience is spread out?
Standard Mandarin is the safe default for a broad audience. If you know a large share of listeners are in Hong Kong or Taiwan, a separate version in their variety is worth the extra work.
Compare all seven voices on the Chinese text to speech page.