search

VOICeVIO / Sinsy

[Fork] Sinsy is an HMM-based singing voice synthesis system.
Other
17 stars 6 forks source link

现在有一个xml文件，怎么生成wav？ #3

Closed leeseean closed 4 years ago

UlyssesWu commented 4 years ago

Sinsy.exe -x dic -m nitech_jp_song070_f001.htsvoice -o song.wav song070_f001_063.xml

leeseean commented 4 years ago

Sinsy.exe -x dic -m nitech_jp_song070_f001.htsvoice -o song.wav song070_f001_063.xml

这个生成的音频声音怪怪的，断断续续，执行命令的时候提示no language，可能我要的是中文或者英文歌词，这个命令只处理日语，所以导致出来的音频断断续续不能曲调

leeseean commented 4 years ago

@UlyssesWu

UlyssesWu commented 4 years ago

Sinsy is developed by Japanese researchers. "Yoko" (nitech_jp_song070_f001.htsvoice) is a Japanese voice model. And dic is a Japanese dictionary.

You can applied it to other languages like Chinese or English with your (self-designed) dictionary and voice models. To make them, you have to gain enough knowledge about HMM & HTS. https://github.com/VOICeVIO/Sinsy/issues/1#issuecomment-553484678 is a good point to start.

There are also some projects trying to make use of HTS/Sinsy in other languages, such as sinsy-remix, Festival etc. You can learn from them to get some inspiration.

However if you have no aspiration for learning a lot (including math, English, signal-processing, coding in C/C++ etc.), Sinsy will be useless to you. And you should just earn enough money to buy Vocaloid / CeVIO/ SynthV etc. Or use some easy tools like UTAU or DeepVocal, which (AFAIC) won't catch the future of Singing Synthesis.

For now I haven't seen any "aspiration" from you. Before you continue, I think you should learn the story of Kanru Hua - the developer of SynthV. Although I'm not 100% with him, he did spent so much on achieving his dream since he was a junior school student. Since then he keeps learning these things (which I doubt you might give up in 3 days) and finally SynthV is there.

Stop asking questions and start learning if you want to be the next Kanru Hua.