Japanese text to speech: AI voices that read Japanese
Japanese is one of the 31 languages of Supertonic 3, the open text-to-speech model behind SuperClonica, so every voice in the catalog reads Japanese: anime and game characters, narrators, male and female voices. You can hear any of them say your own Japanese line for free, and voice long Japanese texts offline on your own computer without a character limit.
This guide explains how to get the best Japanese from it, honestly: how to write kanji and kana, what the engine does with numbers, what accent to expect and how to get a voice recorded in Japanese.
Try any voice in Japanese for free
On the page of any voice, choose Japanese next to the text field, type a phrase of 30 to 100 characters and press Voice it. You hear your words in that voice in a few seconds and can download the recording. No account is needed.
Write Japanese the way it should be read
Short sentences with clear punctuation sound the most natural.
- Write in the usual mix of kanji and kana.
- For an unusual reading of a name or a word, write it in kana: the model cannot guess a reading that the kanji do not show.
- Numbers can stay as digits: the engine spells them as Japanese words before synthesis.
- Use 、 and 。 for pauses and sentence ends; ！ and ？ change the melody of the line.
Anime lines in the original language
Japanese suits anime and game characters: many of them can say a line in the original language, and the anime voice guide shows how to pick one. For a scene with several characters, voice each line separately and join them in a video editor.
What accent to expect
The voices in the catalog were made from English and Russian recordings. They keep their timbre in Japanese, and the accent of the original can be heard, a little for some voices and more for others. Listen to the demo and your own Japanese line before you choose, or get a voice recorded in Japanese.
Get a voice recorded in Japanese
For a native Japanese voice, train one from a clean recording of 5 to 10 seconds in Japanese, your own voice or a speaker who agreed, in the studio on a computer with an NVIDIA graphics card. No graphics card? Order the voice on the exchange: an author trains it from your recording, and you pay only after you accept it.
Unlimited Japanese offline
For long texts, download the voice file and install the free SuperClonica engine: python -m pip install superclonica-cpu. Then voice a Japanese text file with one command: superclonica --text "$(cat japanese.txt)" --voice voice.json --lang ja --output japanese.wav. It runs on an ordinary processor, offline, with no character limit and no monthly fee.