The Japanese reader is simply unusable. It gets most things wrong from the tokenization to the voice generator. If the problem was occasionally using the wrong reading for some kanji that have lots, I would understand, but it goes much deeper than that. For instance, to read the "hi“ in the image below it uses the English pronounciation /haɪ/, instead of the Japanese /çi/, meaning that the text-to-voice software is probably not even set to the proper language. This is honestly so disappointing for a paid app, and if I wasn’t using LingQ for other languages as well I would be unsubscribing.
1 Like
Yes, the furigana is bad / inaccurate too and don’t match the audio often. Some are only as good as what sources are available. I’m sure you’ve seen bad translations on Google Translate, ChatGTP, and DeepL too. I certainly have. It’s definitely better than it used to be when it would just take a word sometimes and make a katakana version of it without explaining it. Still the accuracy of these dicitionaries are a ways off vs widely spoken European languages. Thus we have this situation. What’s more it can be painstaking to customize and update particularly if it is personally uploaded materials which others have not gone over previously.
1 Like