explainroo 0.3: Japanese and Mandarin narration

explainroo 0.3 adds Japanese and Mandarin Chinese, a new look, and a speech check that compares how words sound.

Japanese and Mandarin Chinese

Set "language": "ja" or "zh" in video.json, and Kokoro reads the script with one of its 5 Japanese or 8 Mandarin voices. You write the script as usual, without spaces between words. Cue words, markers and captions work as in every other language.

Japanese and Chinese need Python with misaki installed once. misaki is the pronunciation library Kokoro’s voices were trained with. The voices and languages page has the commands.

Two example videos show it: why we have leap years, in Japanese, and how websites store passwords, in Mandarin.

A speech check that compares sounds

Whisper, the speech model behind the check, often writes a word with other characters that sound the same. The check now compares how words sound: Japanese by their kana reading, and Chinese by pinyin, with Traditional and Simplified characters counted as the same. In English, it now reads years and ordinals the way the voice says them, such as 1900 as “nineteen hundred”.

The editorial look

The sixth look has serif headlines and fine lines on warm paper. It suits essays, history and money. See it in why we have leap years.

Also

• A web player only carries the font characters its video shows, so Japanese and Chinese players stay small.
• explainroo is tested on Linux, macOS and Windows for every change, and math formulas now work on Windows.

The full list is in the release notes on GitHub.