Japanese: Correct Pronunciation using Furigana

Text to speech pronunciation of Japanese is not always correct. One way we can help TTS engines achieve correct pronunciation is by using Furigana hints.

Let's say we have a note which reads レストランは空いています。(the restaurant is empty). Some TTS engines will pronounce this as resutoran wa aiteimasu. The correct, expected pronunciation is resutoran wa suiteimasu.

Let's use the Furigana notation to give a hint to the TTS engine: レストランは 空[す]いています。

In the HyperTTS Text Processing section, we have to add a Regex rule, with the following parameters:

• Pattern: ([^ ]+)\[(.*?)\]
• Replacement: <sub alias="\2">\1</sub>

The input we want to send to the TTS engine is レストランは<sub alias="す">空</sub>いています。You can verify this in the Processed Text column on the grid on the right side. You can select the row, and click Preview Sound to verify that the pronunciation is now correct.

This method requires an SSML service. It should work on the following services:

• Amazon
• Azure
• Cereproc
• Google

Unfortunately, it won't work on non-SSML services such as Naver, VocalWare, Google Translate.

Note: alternatively, you can use the regex ([^ [\]]+)\[(.*?)\], which will support multiple Furigana characters, like in 元[げん]気[き].