Wiki Wednesday #311 - VoiceFace's Cloter

 ouch, my stomach hurts.

Wiki Wednesday #311 - VoiceFace's Cloter

Cloter is a 43 year old man with anxiety.

Who Is Cloter?

Art from site

.

Diffsinger cannot pitch your voice up! Nor can it pitch it down. That's just not how it works, which is super insane for an utau user!

That's not really enough to extrapolate into five-hundred words. If you don't record at a certain pitch, your Diffsinger won't sing at that pitch. You can use an effect like ReaPitch to pitch your samples up or down if you struggle to sing at that pitch. 

So instead... Ever notice how you almost never see people actually using the pitch models? You select the notes and under "batch edits > notes" you choose "load rendered pitch". People don't seem to use that a lot.

In my experience, it's a little wonky. I thought training with "melody encoder" made it better, but all of the pitches kind of look the same when you train with that, so I'm switching back to standard.

There's a program called SlurCutter that you're meant to use to fix the automatically generated MIDI files to improve your pitch model... But it doesn't seem really important to do so.

I can't throw shade, though. All of the tester Japanese USTs I have all have pitch bends that I don't know how to remove in OpenUtau. So when I test a Japanese bank, it's going to have the pitch bends I got from UtaLis. Heck, even when I was testing English banks, I got tired of rendering pitch and just kept the same pitch from the last bank I had bothered rendering it for!

Training a pitch model is really useful. Having one good one that you use on all of your USTs as a base is a pretty good idea! You're never actually forced to use your model's pitch model. If you load a different model and load rendered pitch, you can then switch back to your model and have that super nice AI generated pitch from that other model with no issue. As far as I'm aware, you could also slap an UTAU onto that UST and keep the cool AI pitch bending.

As a note, loading rendered pitch may actually not be fully possible for you. Sometimes, your computer will decide it has had enough and will give you an error instead of giving you the pitch lines that you wanted! In this case, manually tuning is your only option.

So the pitch model is really cool. I actually really dislike tuning, so it's totally a necessity for me! But you don't really need to worry about it. I wouldn't really recommend anyone actually going through and fixing the automatically generated midis unless it was really important to them personally. Besides, the midi generation seems to be improving with time. I heard someone talking up SOME for analysis and I've been using it in the Colab ever since with no issues that I notice.

People in the vocal synth fandom just really, really like drawing their own pitches. I get it, even if I personally don't like doing all of that.

How Is Cloter's bank?

Cloter originally had a different voice provider. Likely to cut down confusion, his current one pitch CV bank has the subtitle of "Rory". He has a very nice and masculine voice!

Where can I download Cloter?

You can find him on his official site. He is cool!