mirror of
https://github.com/zarzet/SpotiFLAC-Mobile.git
synced 2026-11-21 20:00:49 +01:00
Lyrics in other alphabets now show how they are sung without an AI key or a network: Korean (Revised Romanization with liaison, nasalization, lateralization, aspiration and palatalization), Cyrillic with one standard per language (Russian, Ukrainian, Belarusian, Bulgarian, Serbian, Macedonian, Kyrgyz, picked from the letters of the song), Hindi and Punjabi in the plain lyrics spelling with dropped inherent vowels, Chinese Hanyu Pinyin with phrase readings for polyphones, and Japanese kana by Hepburn. Japanese lines with kanji are not romanized offline: a reading dictionary does not fit the size budget. Those lines, and scripts the device cannot romanize (Thai, Arabic and others), are what an AI translation is asked to romanize when it is set up to; otherwise they show no romanization. A translation no longer asks the AI to romanize what the device can. Pinyin comes from a 259 KB gzipped asset built by scripts/build_pinyin_data.dart from Unihan (Unicode License v3) and phrase-pinyin-data (MIT); both notices are in the app's licenses. It loads on a background isolate the first time Chinese lyrics play. Word-timed lines keep their timing: the line romanizes as a whole, so sounds that run across words read right, and each word takes its own part. The romanization is shown, never saved with the song. Lyrics settings get Romanize lyrics (it starts as the AI romanization switch was, so whoever turned that off is not shown romanization) and Romanization as main line, which shows the romanized text as the line with the original under it; a word-timed line whose romanization has no timing keeps its original as the main line.