Dunder V2 from Z Vocal Project released!

Hello everyone! We are rounding out the year with the release of “Dunder V2 from Z Vocal Project”, a massive update to Dunder’s English CoeFont voicebank.

Synthesis quality has improved greatly over what is now called “Dunder V1 from Z Vocal Project”, containing maybe about an hour worth of recordings. For the time being we have decided to keep V1 active alongside V2, as although V2 is superior, he has a completely different tone to V1.
Use V2 here.
Use V1 here.

Dunder V2 from Z Vocal Project 15s Commercial
A New Dunder Design

Dunder’s new design for his CoeFont V2 is an inversion of his usual colors. I think he kind of looks like a British Exchange Student now, so this design is being called his British Design. Despite the name, Dunder V2 is still American Accented, this is just what he might look like if he ever went to the UK for an extended period of time and then came back to Delta.


Comparison with V1

The phrase is “Hello. I’m Dunder from Z Vocal Project. It’s nice to meet you”. Dunder V1 is on the left, Dunder V2 is on the right (If on mobile V2 will be below probably?)

As you probably heard, Dunder V1 has a much more solid tone whilst V2 is a a bit on the softer side. This was tone with the intent of making him more consistent for narration, as the point of this product is to aid in accessibility more than provide a character vocal. Dunder V1 was originally recorded with a tone similar to Dunder Announcer from Z Idol Talker.

We’re not done with Z Idol Talker!

Speaking of Z Idol Talker, we have not stopped development, but are simply expanding the platforms to launch voicebanks on. I’m actually working on three voicebanks right now. Dunder Chinese, Dunder Polish, and Jay★Monotoloid Chinese. These voicebanks need a bit more work, as opposed to other language appends, the training data for these voicebanks were generated via ElevenLabs. Please listen to some samples if you’re curious.

Dunder Chinese
Dunder Polish
Jay★Monotoloid Chinese
Thanks for reading!

I’m currently debating whether or not to release the autolabeled Diffsinger voicebank I trained earlier this month as an alpha, as it has a lot of issues, and too many Dunder voicebanks are in alpha/beta right now.
I’ve attached a demo down below for the time being. This is his Hushed voice color. I can’t sing below like G2 just yet and it seems like DiffSinger cannot handle going below the training range like ENUNU could.

Early Alpha of a Dunder Diffsinger voicebank singing an English Cover of Someone Else’s Man

Also, depending on how life goes for me in the coming months, I might be looking to find Zra a VA.

Talk to you guys again soon hopefully!