Speech Synthesis in SpeakGoodChinese

How to change the pinyin synthesis

Tone and Speech synthesis in SpeakGoodChinese

SpeakGoodChinese has three levels of tone and speech synthesis beside the option to add real spoken examples (Real examples button). Which synthesis method is used is choosen on the Settings page:

The "Humming" option will stress the tone realizations. The two TTS options will also practice pronunciation, or at least phoneme perception. The phoneme pronunciation of the native Windows and Mac TTS is much better than the build-in, rule-based, eSpeak synthesis. However, the tone realizations of the high quality native TTS are probably suboptimal for beginning students trying to practice tones. It is probably best when students switch to the native synthesis on Mac or Windows only when they have reached a basic level of tone perception. So, the prefered order is likely for students to start with the build-in eSpeak voices or even the humming synthesis.

Text-To-Speech synthesis in SpeakGoodChinese

SpeakGoodChinese uses the eSpeak Text-To-Speech (TTS) engine that is build into Praat. eSpeak has several Mandarin voices. You cycle through them when you click on the Voice button on the Settings page. Voices that start with an F are supposed to be "female" voices, those starting with an M are supposed to be "male" voices.

The voice labled "*" is the system's default synthesizer. On a Macintosh computer, this is the Say program. On Linux this will be the spd-say speech dispatcher, if it is installed. The Macintosh Say program has a rather good synthesis quality. The Linux spd-say program is more like another eSpeak instance.

Installing Chinese voices on the Macintosh (OSX)

If the "*" voice does not work on your Macintosh computer, you probably do not have the correct Mandarin voice installed. To install Mandarin speech synthesis on you Macintosh, open System Preferences.... Locate and open the Dictation & Speech item. Select the Text to Speech tab. Click on the System voice menu. At the bottom of the System voice menu, select Customize... and scroll down until you find Chinese (China) and tick the box for Ting-Ting.

Your computer will download and install the selected voices. This can take a lot of time.

Installing Chinese voices on Windows

If the "*" voice does not work on your Windows computer, you probably do not have the correct Mandarin voice installed. The installation process is different for Windows 7, 8, and 10. Please refer to the procedure for your system. When you select a Mandarin Chinese voice, make sure you install the simplified characters and a Chinese zh-CN voice! SpeakGoodChinese uses a female voice. SpeakGoodChinese is known to work with the Huihui voice.

Changing to another speech synthesis program

SpeakGoodChinese can use external Text-To-Speech synthesis programs as long as they have a command-line interface (CLI). The command to call this program on the computer must be stored in a text file in the TTS folder in the Preferences directory.

Note: This has not yet been tested on Windows.

First, locate the Preferences directory on the computer.

For a single user <User name>, the local Preferences directory is:

To install this for all users, place them in the global Preferences directory:

The name of the text file containing the TTS command depends on the computer platform:

The command file should only contain a single line of text with the command to speak the text. The commands to call the TTS synthesis should be constructed in such a way that a pinyin string can be appended at the end. Note that the neutral tone will be indicated by the numeral 5.

Often the (pinyin) words need to be placed at another position in the command string or the command even needs Chinese characters or expanded pinyin. Then the place of the pinyin or text should be marked by a placeholder:

Examples of such commands are:

While such a file is present for the operating system running SpeakGoodChinese, the program will only use that to synthesize speech.

Links to this page

© R.J.J.H. van Son, September 16, 2015