Trem's Demo! (Fly me to the moon)--English

" frameborder="0" loading="lazy" allowfullscreen >

It took me a bit to finish this; I kept replaying the file in audacity just to listen to it, lol.

I added some effects to get that muffled sound, btw, she does sound clearer than that after export from UTAU. I kind of had fun with the whole "1930's recording" feel.

I had fun with pitch bending at the end, too. And vibrato. This is the first time I've used vibrato, acutally...

All right, I'll stop babbling. Thoughts?

I think it sounds nice!

smeen wrote:I think it sounds nice!

Thank you! :D

Woww, it surprised me how understandable she is! Was it difficult to get right? I'm reluctant to make an english voicebank because there's so many sounds and it's not usually that understandable DX

She does sound nice x3 GJ!

xMienaiTsubasa wrote:Was it difficult to get right? I'm reluctant to make an english voicebank because there's so many sounds and it's not usually that understandable DX

She does sound nice x3 GJ!

Thank you :]

It actually wasn't that many sounds, I made this with a 333-sound voicebank.

This song in particular wasn't difficult to make, but I know it was really frustrating for me the first time I ever tried english. I think it's the fact that you have to think of the program differenly than you do for japanese. Not everything is the 'consonant-vowel' formula, so I had to get inventive sometimes. And use your ear a lot to gauge things, which means listening to stuff repeatedly a LOT.

Blending is the key to make english sound decent, I think--for example, to get Trem to sing 'and' is actually three sounds; 'aah', 'nn', and a 'd' sound (I used 'da' because the 'd' is separated a bit from the 'a' in my recording). 'nn' has an overlap on 'aah' of about 150, and almost nothing after that (which makes it hard to highlight, lol). Then the 'da' sound is shortened to where you just hear 'd', and you stick it on the end.

Audacity helps, too, 'cause I can cut out gaps that sound unnatural. Or if Trem's taking too long on the 'n' part of 'na', I'll just hack out part of the 'n' sound and make sure it doesn't make it sound choppy.



Wow, I just wrote a book. lol.

That was really useful, actually XD
So to get the sounds to blend and not sound choppy, you have to play with the overlap and stuff a lot?

Hm, I'll definitely have a go at using Trem's bank when she's finished XD

xMienaiTsubasa wrote:So to get the sounds to blend and not sound choppy, you have to play with the overlap and stuff a lot?

Yeah, I just figured it out by experimenting with overlapping. Every word is different. Although I've found that I can copy and paste an 'n' sound from the end of one word and put it on another and it'll sound fine.

Also, you have to think about stringing sounds together differently too; like the word 'play'. If you put it together using 'p' 'll' and 'ay', it's going to sound choppy and you're going to have a really hard time getting it to sound right. But if you stick a 'p' sound in front of a 'lae' sound, it sounds much more natural. You have to think about where the sounds are grouped when you say them.

Also, there's a lot of tricks that actual singers use to make english sound nice that you can use for UTAU, too. One of the biggest rules that will help a lot is this:

Focus on the vowels, and minimize the consonants to the beginning and end of the word.

In other words, you don't say 'nnnnow' when you're singing; you enunciate an 'n' so that people can hear, and then you open out to the 'ao' part, holding on to the 'a' for as long as you can before you tack on the 'o' sound at the end. Also, you keep the vowel sound as constant as you can, only changing when you come to the end of the word. Which is kind of a 'duh' in UTAU, 'cause they can't change their wav files, lol.

Oh, and always make 'k' and hard 'c' sounds crisp; too many times I've heard UTAU's say "khhhaaa" when it should be a nice, crisp "ka" sound, and it detracts from the singing 'cause it sounds like their tongue suddenly got stuck in the back of their throat, or they can't breathe right, or something. X(

(Trempush<--former choir nerd) I hope I made sense, there... O.o

xMienaiTsubasa wrote:Hm, I'll definitely have a go at using Trem's bank when she's finished XD

8D Yay! I'll get it up soon :D

Her voice is clear and her English is understandable. This = Awesomeness *3*

moeloid wrote:Her voice is clear and her English is understandable. This = Awesomeness *3*

thank you! :D

Some of the other english banks I've heard were a little mushy, but Trem's is extremely crisp and clear. Love it~ Love Sinatra. Love Eva. Love Trem. Can't wait till she's finished. :D And yeah, I like ze mode2 functions too. xD

KURENAI wrote:Some of the other english banks I've heard were a little mushy, but Trem's is extremely crisp and clear. Love it~ Love Sinatra. Love Eva. Love Trem. Can't wait till she's finished. :D And yeah, I like ze mode2 functions too. xD

Thank you! This makes me so happy that people like her! :D