bryanjenningz
commented
3 months ago The audio is made using the text-to-speech web API, here is the code that uses text-to-speech.
The reason why the text-to-speech web API has errors with pronouncing "開く" as "ひらく" instead of "あく" is because the code is passing in the kanji instead of the hiragana. If we pass in the hiragana instead of kanji, then the pronunciation will match the hiragana. However, then we will have more pitch accent errors because we lose pitch accent information if we only pass in hiragana.
A solution would be to add a settings option which can be toggled to play the hiragana instead of the kanji if a user cares more about the hiragana being played than the pitch accent being correct.
Another solution would be to change the hiragana so the text-to-speech of the kanji matches the hiragana.
Another solution would be to just remove words where the text-to-speech of the kanji doesn't match the hiragana.
This would take a lot of time since there are thousands of words and the text-to-speech was relatively accurate so I decided to just not do anything.
I also agree that the styling could probably be improved a little bit :)
Adding sentences would be good so the words could be learned in context but this might take a while since there are thousands of words. Maybe a simple solution to this would be to use an API which provides example sentences for each word.
FragozoLeonardo
Hello, your app is a awesome idea, but I have a few points on your application.
is the audio narrated by a human or it's TTS? another: some words have mismatched audio eg: are 開く (read as あく) in the audio being told ひらく, a whole different word.
Other than your application is really nice, but I think that the styles can be more... spaced I guess.
Other than I would add sentences as well.
Kind regards.