Dunder_Intercalaris AI now in development for Enunu!

Sorry for not posting last month, it was a bit hectic, but I bring some exciting news with me! We have decided to restart development of Dunder’s Japanese Enunu voicebank, officially giving it the name Dunder_Intercalaris AI. We aim to have it completed for his 6th anniversary.

Key art

New Design

Dunder has been given a new design just for the occasion, sporting baggy clothing and a bag with an OpenUtau keychain attached to it. For the hell of it I also drew the design over Mazata’s Dunder model, so you also get Tiny Dunder_Intercalaris AI. If I see anything weird done with him I’m taking him away.

The Intercalaris logo has also been decided upon for the time being, meaning a new round of logos will be posted after this post goes up. This includes the logos for the unreleased Dunder_Intercalaris AI and Dunder_Intercalaris+. Feedback is appreciated as we cannot afford professional logo design at the moment.

Dunder AI Samples

With that out of the way it is time to showcase some of the first samples of Dunder_Intercalaris AI. As a reminder I cannot sing and this voicebank is only going to be intended for use with OpenUtau as it supports manual pitch editing. I will showcase what his auto-pitch sounds like right now for the hell of it, but it is not good, and I do not expect it to get better.

First up is the worst of all the samples, an untuned Japanese cover of my song Hard Drive. There’s a lot of “sh” in this song and that’s one of the biggest struggles of the voicebank right now, alongside adding breaths where they don’t need to be cause I think the guide I was following just lied about not labeling breaths as breaths.

Hard Drive (Japanese Version) – Dunder-P

Next is a sample of Fake Me by Kanako Itou. One of the samples is gonna be no tuning, the other is going to be the automatic tuning. As you will soon see, it is a better idea to just stick to manual tuning.

Fake Me – No Tuning – Kanako Itou
Fake Me – Auto Tuning – Kanako Itou

Finally, is a tuned sample of Koi by Gen Hoshino. One of the samples is SynthV Auto-Pitch tuning, the other is the voicebank’s tuning. I’m gonna label which is which but like it’s immediately going to be obvious.

Koi – SV Tuning – Gen Hoshino
Koi – Voicebank Tuning – Gen Hoshino

The difference between this voicebank tuned and untuned is like night and day but none of my SVPs will convert correctly and I don’t wanna use someone else’s UST so this is the only real example I can give of this.

The plan going forward

The plan for Dunder_Intercalaris AI over the next week or so is to record and label more data focusing on his high range, along with relabeling the current data to fix the breath issues. Following that more data will be collected in general until we reach at least an hour or so, as it’s only at 20 minutes counting silences, and then labeled, and if he sounds okay after that, we will start flagging tones. In the event this gets done before June 11th we will probably finish up Dunder_Intercalaris+ and potentially even start on a Dunder_Mercedonius AI. If Dunder_Intercalaris AI is not finished before June 11th we will release a Beta voicebank on that date just to make sure there is still something out to mark the 6th anniversary.

Mercedonius out than finish Intercalaris if I’m being entirely honest.

Thanks for reading!

This concludes this article. Thank you for catching up with the ongoings of Z Vocal Project. I’m not entirely sure if there will be another post this month, but if there is it’ll have art of Dunder, Kiyana, and Zra together for Black History Month.
Oh also the website was hard to get into for a few days huh. I’m gonna upgrade the ram soon, this should happen less often in the future, sorry for the inconvenience.