I'm having trouble with making Saki say "dream" in Japanese accent.

Anyone know the trick to make her say “dream” in Japanese ?
image_2022-10-31_195253572

You say with a Japanese accent, are you using cross-lingual synthesis to sing in English, or do you have Japanese selected? The answer will be different based on which phoneme list you are using.

In Japanese, “dream” would be pronounced 「ドリーム」(romaji: “doriimu”). So the phonemes would be as follows:
image

Of course you can adjust the timings to suit your song.

In English arpabet, to mimic the same accent, you might try something like the following (though this is an approximation, so don’t be afraid to try a few variations):

image

1 Like

The song is sing by Aimer (Japanese song) and she pronounce dream just like English.
I’m trying to use Japanese to sing that word. So the notes has to be pretty short to say that word?
This is the song at that current time when she say “dream” Aimer - Ninelie - YouTube
At the moment I’m using multiple tracks to make Saki pronounce “dream” and continue to sing Japanese seamlessly, but I think there might be other way because editing multiple track at the same time is making me pretty confusing.

Ahh I understand. You have two main options:

  1. Instead of using two tracks, you can embed a Group in your main track that has English selected. Right click on the “dream” note (or both “day” and “dream”) and select “Merge into group”, then select English for the group. Groups and Tracks are very similiar in SynthV, so using a Group like this is sort of like nesting one Track within the other. That said, it’s still not a seamless transition because the group is rendered separately from the track it is part of, just like two separate tracks would be.
  2. Find some Japanese phoneme combination that sounds close enough to Aimer’s pronounciation. This will probably be tough, since she uses near-fluent pronounciation of those English words, and the Japanese phoneme list simply doesn’t have a sound to match the English r. If you want to experiment with this approach, you can try removing or shortening “excess” phonemes (ie d r i m instead of d o r i i m u), though removing them entirely might result in slurred or unclear sounds due to consonant transitions that don’t exist in normal Japanese speech. Another technique is to find the nearest similar sound, for example when spoken quickly “dream” can sound a lot like “jream”, which might give better results depending on the specific voice database and word.
2 Likes

Thank you very much!! I will try it and see if it work.