So, I had a random idea...

[ReticentResolve]

I just thought of this when I was looking at the oto for my arpasing bank. Just a thought, but UTAU has it so that if you make a CV sample, the beginning of it will play no matter what, right? No matter the length of the note (So long as you oto it). Would it not be able to support doing the same to the back end of the note? So that if you sing 'What', you can have a sort of reverse CV where it sings 'ah', and always stops with a 't'?

Is this actually possible? Or am I just being crazy?

 

[Kiyoteru]

No.

Edit: Wait a moment, this post had nothing to do with website feedback. Would someone mind moving it?

 

[HoneyPai]

Technically yes it's possible but it probably, more than likely, will not sound good

 

[partial]

Wouldn't this.....just be CVVC?

If you're asking if you can reverse an OTO, such as by having the pink(consonant) covering the end t sound as well as the silence after it, and the white being before the t, the answer is no. (However, if this were possible it would be really useful I think!!)

Here's a crappy diagram of what I mean.

 

[FeatheredFinch]

Wouldn't this.....just be CVVC?

If you're asking if you can reverse an OTO, such as by having the pink(consonant) covering the end t sound as well as the silence after it, and the white being before the t, the answer is no. (However, if this were possible it would be really useful I think!!)

Here's a crappy diagram of what I mean.
View attachment 5291

Someone needs to mod Utau so this is possible someday

 

[Damien]

Wouldn't this.....just be CVVC?

If you're asking if you can reverse an OTO, such as by having the pink(consonant) covering the end t sound as well as the silence after it, and the white being before the t, the answer is no. (However, if this were possible it would be really useful I think!!)

Here's a crappy diagram of what I mean.
View attachment 5291

We like, need this.

 

[FeatheredFinch]

We like, need this.

This would make other languages so much easier, we need to kidnap ameya and force her to update Utau for us, whos in?

 

[KNΞMΛTCS]

This is something that would really benefit CVVC banks for sure. I think Moresampler should implement it - you could just have an insanely high value for one oto parameter to act as a trigger to tell it that it's a VC and do what you want it to.

Edit: I'm specifically referring to handling the "backwards" VC right - stretch the beginning and leave the end alone. I don't know so much about OPs idea.

 

[na4a4a]

This would not sound good the way you explain it.

We have CVVC which is good enough for all languages and is the best method for anything complex.

Always landing on the consonant is not even a feature we need, just use the silence after it and it will always land where you want it.
The oto already lets us do this.
preutterence will place the consonant at the start of the note and that's it, it will always sound.

 

[VocAddict]

What would be the point actually? For example, the word 'what', you would have [w ah] [ah t] where the first note would be longer, stretching the vowel, and you'll have the short VC at the end for the consonant at the end. What difference would it be to stretch the vowel of the VC? Like, you can even do [w ah] [ah] [ah t] if you so desire.

There's honestly no reason why I will think this would be necessary. What benefit would it really bring? Also, "modding" UTAU to support this will basically force all tools used for it right now to basically mess up. Wavtools, resamplers, plugins, setparam, basically everything would have to update just to support this new feature.

I believe the system we have in place right now does a perfectly nice job of handling VCs even though it might be considered ancient.

 

[KNΞMΛTCS]

What would be the point actually? For example, the word 'what', you would have [w ah] [ah t] where the first note would be longer, stretching the vowel, and you'll have the short VC at the end for the consonant at the end. What difference would it be to stretch the vowel of the VC? Like, you can even do [w ah] [ah] [ah t] if you so desire.

There's honestly no reason why I will think this would be necessary. What benefit would it really bring? Also, "modding" UTAU to support this will basically force all tools used for it right now to basically mess up. Wavtools, resamplers, plugins, setparam, basically everything would have to update just to support this new feature.

I believe the system we have in place right now does a perfectly nice job of handling VCs even though it might be considered ancient.

It gets the job done, but there's always room for improvement. This idea might give smoother phoneme blends, although I think the real key to that is analyzing the volume of the samples and having the resampler/wavtool do an advanced envelope-style edit, creating a completely smooth blend with no work from the user.

 

[VocAddict]

It gets the job done, but there's always room for improvement. This idea might give smoother phoneme blends, although I think the real key to that is analyzing the volume of the samples and having the resampler/wavtool do an advanced envelope-style edit, creating a completely smooth blend with no work from the user.

This is something I can see moresampler supporting something like this in the future. One thing that always bugged me was that envelopes in UTAU affects volume of the entire sample itself and not just where the the crossfading takes place.

 

[Kiyoteru]

I'm going to entertain this idea a little bit, even though I don't agree with it and frankly think that this is only being brought up because people would rather see UTAU change entirely than learn how to properly configure a VC.

The least chaotic way to introduce such a change would be by leaving the format of the OTO almost exactly as it is. This would ensure maximum compatibility with existing tools. I think it could be notated through negative consonant value.

Normal
ka.wav=ka,100,300,-600,60,30

Reverse
ak.wav=ak,100,-100,-200,150,50

(I realize that the exact placement of the parameters in this diagram is not ideal, however, I'm just illustrating the idea.)

All other parameters are the same, but the fixed region is now measured relative to the cutoff instead of relative to the offset.
UTAU's default oto editor would probably freak out, and lots of resamplers would freak out too, but newer resamplers under active development could account for reading a negative consonant and processing it appropriately. Maybe SetParam could see updates (and iirc it might be open source?) in order to edit otos like this.

EDIT: I think I'll page @kimchi-tan to move this thread to UTAU Discussion.

 

[ReticentResolve]

I'm going to entertain this idea a little bit, even though I don't agree with it and frankly think that this is only being brought up because people would rather see UTAU change entirely than learn how to properly configure a VC.

The least chaotic way to introduce such a change would be by leaving the format of the OTO almost exactly as it is. This would ensure maximum compatibility with existing tools. I think it could be notated through negative consonant value.

Normal
ka.wav=ka,100,300,-600,60,30

Reverse
ak.wav=ak,100,-100,-200,150,50

(I realize that the exact placement of the parameters in this diagram is not ideal, however, I'm just illustrating the idea.)

All other parameters are the same, but the fixed region is now measured relative to the cutoff instead of relative to the offset.
UTAU's default oto editor would probably freak out, and lots of resamplers would freak out too, but newer resamplers under active development could account for reading a negative consonant and processing it appropriately. Maybe SetParam could see updates (and iirc it might be open source?) in order to edit otos like this.

EDIT: I think I'll page @kimchi-tan to move this thread to UTAU Discussion.

Sorry, I wasn't sure where to put this. And I do actually know how to oto VC, one of the reasons I thought of this was because when I input the phrase into UTAU, i.e [w ah] [ah t], there is always a seperation between them, and you can obviously tell where one begins and one ends.

I know this is likely just an issue with the bank, and it can be mostly fixed with enveloping. This was just an idea.

 

[kimchi-tan]

No.

Edit: Wait a moment, this post had nothing to do with website feedback. Would someone mind moving it?

Moved!

 

[Kanru Hua]

It's certainly possible in the general sense of doing speech/singing synth, but unfortunately not in UTAU, as the software is "hardwired" for Japanese. The problem you'll run into is that the vowel preceding the consonant can't be time-stretched. This had been a major limiting factor in the design and implementation of Arpasing.

 

[KNΞMΛTCS]

Kanru Sama has arrived OTL

That's quite unfortunate to hear.