Dread

So I'm currently working on a sampler that allows you to control the mouth position of a voice sample, particularly a creature voice. My current process is to have many different mouth position loops recorded, and have a controller to blend between these samples naturally. Now, the issue is that standard volume crossfading doesn't work with vocals, as instead of having a natural blend, the center of the fade sounds like 2 separate voices playing at the same time (which is what that is).

I do have some other solutions to this, such as recording the transition of the mouth positions and triggering them when you wish to move, for instance if you are playing an A sample, you can then trigger it to play an A to O, which would then blend in with an O sample. This does work, but it limits the amount of control you have, and how quickly the transitions can occur. It also makes it more challenging for users to load in their own samples which is something I want them to be able to do.

It's a really odd and stupid sounding question, but does anyone know if there is a way to crossfade that allows the samples to blend into each other in a way that doesn't sound like 2 separate samples? I don't have much faith that this will work but its worth an ask before I give up.

So I'm currently working on a sampler that allows you to control the mouth position of a voice sample, particularly a creature voice. My current process is to have many different mouth position loops recorded, and have a controller to blend between these samples naturally. Now, the issue is that standard volume crossfading doesn't work with vocals, as instead of having a natural blend, the center of the fade sounds like 2 separate voices playing at the same time (which is what that is). 

I do have some other solutions to this, such as recording the transition of the mouth positions and triggering them when you wish to move, for instance if you are playing an A sample, you can then trigger it to play an A to O, which would then blend in with an O sample. This does work, but it limits the amount of control you have, and how quickly the transitions can occur. It also makes it more challenging for users to load in their own samples which is something I want them to be able to do. 

It's a really odd and stupid sounding question, but does anyone know if there is a way to crossfade that allows the samples to blend into each other in a way that doesn't sound like 2 separate samples? I don't have much faith that this will work but its worth an ask before I give up. 

So I'm currently working on a sampler that allows you to control the mouth position of a voice sample, particularly a creature voice. My cur

vocal-crossfader

you want to crossfade between 2 sources but without crossfading between them. 

that is a wide field and i am afraid there is no simple algorithm which finds intermediate steps between A and O for you - it would be simpler to synthsize it from the beginning.

not to speak of that "doubletracking" issue you already correctly pointed out.

so as a start my first suggestion would be to use 5 or 8 different recordings for the transition.


you want to crossfade between 2 sources but without crossfading between them. 

that is a wide field and i am afraid there is no simple algorithm which finds intermediate steps between A and O for you - it would be simpler to synthsize it from the beginning.

not to speak of that "doubletracking" issue you already correctly pointed out.

so as a start my first suggestion would be to use 5 or 8 different recordings for the transition.

Yeah I thought it was a bit of stretch, I will try out synthetic vocals in the future, but for now I'll try to make an easy to use interface for a large amount of quick transitions.

i won't be able to guide you on how to implement it in a patch, but generally speaking, for the crossfade portion you could use some form of FFT processing to 'morph between the timbre' of each voice(it's called 'spectral interpolation' or 'spectral morphing')... as an example, you could try it out in a program like audiosculpt, or even metasynth(capture spectrogram pics of the vowels and morph between the pics... i think aphex twin did something like this for the song, 'windowlicker' if i recall correctly)... anyways, just something to look into... it can also come out sounding synthetic since it involves a resynthesis stage, but... could be a powerful enough transformation for what you need.

does anyone know if there is a way to crossfade that allows the samples  to blend into each other in a way that doesn't sound like 2 separate  samples?

Yeah I think that might be worth a deeper look for me. I had tried a basic convolution but it ended up making a very weird noise at the midway blend point, but I'm expecting I'd have to go a bit deeper to get what I'm after. Cheers!

Vocal Crossfader