Hello! The Japanese r is very different than the english r, it is in between an english r and l. Depending on accent, dialect, and/or context it may sound more r or l.
Vocaloid also lacks context sensitivity that machine learned and advanced tts systems have. While Vocaloid tech is really cool on my opinion, it is not accurate in terms of automatic pronunciation all the time.
Last edit: Avinash Persaud 2021-02-09
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I believe there are some pronunciation mistakes with your Sinsy.jp service, such as:
ら り る れ ろ
Rarirure ro
Sinsy: https://clyp.it/ec0a2nzs?token=7ada72c989cbda03819bfb45e1a17111 VS Vocaloid 5: https://clyp.it/remepxjk?token=9630355011da4ecb4c7f17e52ca01fd3
MusicXML sample: https://raw.githubusercontent.com/AndroYD84/Files/master/JaBug/rarirurero.musicxml
Shouldn't it sound more like "ra ri ru re ro" rather than "la li lu le lo"? (Here I emphasize the difference between RE: https://clyp.it/gmnvofiz?token=97b3e84220991e7c450d7f7a7148c1ea and LE: https://clyp.it/5e1jj5wt?token=7f1d9990fa471275cf7341049cbc73b4 )
Happens as well with these lyrics:
お く す り のん で え ら か っ た あ し た げん き に なあ れ
O ku suri no nde e ra ka tta a shita gen ki ni nāre
Sinsy: https://clyp.it/zxdz4teg?token=9d0a6530d4fa04a5b7038c489c4efe24
MusicXML sample: https://raw.githubusercontent.com/AndroYD84/Files/master/JaBug/hsong.musicxml
On the last part pronounces "nā-le" instead of "nā-re".
I don't speak japanese so there might be a reason or a rule that I'm not aware of, forgive me if I said anything stupid.
Hello! The Japanese r is very different than the english r, it is in between an english r and l. Depending on accent, dialect, and/or context it may sound more r or l.
Vocaloid also lacks context sensitivity that machine learned and advanced tts systems have. While Vocaloid tech is really cool on my opinion, it is not accurate in terms of automatic pronunciation all the time.
Last edit: Avinash Persaud 2021-02-09