If you read the last article you may be wonder how the hell Dunder_Mercedonius AI has samples already. Simply put while waiting for Dunder_Intercalaris AI to train I managed to record an hour of English data and then stumbled upon PixPrucer’s autolabeling notebook, and after bulk renaming phonemes in Notepad++ I was just immediately able to start training the voicebank.
Voicebank Info
Dunder_Mercedonius AI Alpha is about an hour and 20 minutes of pure English singing data provided by yours truly. Do to it just being straight automatic labeling Dunder_Mercedonius AI has been slapped with an Alpha label and will be given the Beta label when I label the tones and also get some the breaths labeled that I don’t think the notebook would’ve done.
As with Dunder_Intercalaris AI, there was no pretrained model used or anything like that. I will mention here that Mercedonius AI uses Intunist’s English Header as I don’t know when I’ll be adding these to the itch.io page where this will be mentioned.
As it is forbidden by the license for Intunist’s English Header, Mercedonius AI will not be covered by the upcoming Dunder commercial license and thus, do not plan to use this voicebank commercially, their JP Header doesn’t seem to have the same restriction so assumedly Intercalaris AI should be good.
Samples
The following are just some test samples with some of (mostly) my SVPs since they’re almost 1:1 with OpenUtau, you’ll probably hear him fully mixed and tuned and all that when the voicebank actually releases.
First up is the very first test of the voicebank, an English cover of Someone Else’s Man. This was just a JP SVP with English lyrics hastily pasted in
Next is “A Way of Life” (The P3D remix), but it’s missing a section cause I didn’t want to render all 5 tracks of my SVP.
I don’t have that many English SVPs so here is “Naming You” (SVP by K3RA)
This is an extremely old SVP of mine but Road Less Taken
And then finally in terms of English samples, School Days, so you can hear what he sounds like alongside Intercalaris AI Beta. Intercalaris AI’s consonants improved somewhat but also still sometimes bug out, which is why it still has the Beta label.
Finally, I followed through on pitching up and autotuning Dunder_Intercalaris AI’s samples and training Ascenia AI. This voicebank is going to be called Ascenia.NDS, it’s not formant shifted or anything, so by pitching it down an octave you essentially just have Dunder_Intercalaris AI again as the autotune didn’t actually effect anything. If I do the same for the English bank I will try formant shifting just for the sake of them actually sounding different pitched down like actual Ascenia. UST by xMienaiTsubasa
New Design
I uploaded him to the site yesterday, along with posting an OpenUtau screenshot on Twitter, so you may have already seen it, but I’ve done the art for Dunder_Mercedonius AI! This is his “Formal” design and is basically just a pimped up version of the school uniform cause I thought the colors should be the same. I’ll have to make a reference later to actually show off his Jacket but that’s like, the primary difference, alongside the OpenUTAU earing.
Alpha/Beta Release Info
Dunder_Intercalaris AI Beta and Dunder_Mercedonius AI Alpha will be released by the end of the month unless something major gets in the way. I need to make demos for them, get Mercedonius’s logo done, and then prep their voicebanks for release on Itch.io. Ascenia.NDS will release alongside them, but will probably not get a public demo.
The Commercial Dunder License will likely release on or around his 6th anniversary unless I am informed that the Japanese header can’t be used commercially. The license will cover: Dunder_Intercalaris AI, Dunder_Intercalaris, Dunder_Intercalaris+ (If it has been released), and any released Ascenia voicebanks at that point, it will also likely cover any free to download Dunder/Ascenia voicebanks released in the future that I’m allowed to sell a license for. If there is literally any demand I will make a physical
Character Updates
I’ve started doing some of the art for the other Z Music Club members I’ve been putting off. So far it’s just Lain’s song Design the updated School Uniform Art for Fuuka and Dawn. I decided rain gets a big trench coat cause he was partly inspired by this fucking kid who rode my bus who started doing the same thing in high school and I realized it’d help make him a bit more unsettling.
I had been very hesitant on deciding on birthdays for characters before the game or their voicebank launches, however, I was able to uncover the true release dates of Ascenia, Lain, and Dawn, and what was probably Herculo’s true release date. Because of this I’ve decided to canonize their birthdays along with Dunder’s. From now on:
- Dunder – June 11th (Same as mine)
- Herald – October 29th
- Ascenia – November 26th (Canon date of creation)
- Lain – December 11th
- Dawn – December 11th
As it’ll have its own voicebank soon, Ascenia also has its own page on the site now.
Site Updates
I’m still scared to update the theme on this site to one with a sidebar, but I realized my solution of “Make everything full page width” was really ugly so I’m working on making everything wide width in the mean time.
I’ve condensed things into accordions basically everywhere I could think of to better organize everything. Voicebank descriptions, demos, Character Art, and logos are now neatly compacted and organized instead of you needing to scroll through them all to find what you’re looking for. I’ve also added voicebank information on Herald, Lain and Dawn’s retired voicebanks, Ascenia’s retired and upcoming voicebanks, and Dunder’s private voicebanks, for the sake of buffing out some of the site’s history that got lost in the move over.
The About Us page was overhauled, now featuring a full list of out Social media and a brief FAQ about the company, amongst other stuff.
The News page also finally got a sidebar that lets you view articles in certain categories and with certain tags. I had to manually go and do the categories cause I didn’t know they were different from tags so articles from before today may be miscategorized.
As briefly touched on earlier, we have also done the unthinkable, and tweeted for the first time in over a year.
Thanks for Reading!
Wow that was a lot to cover. I hope you look forward to this voicebank’s release. Having a decent sounding Dunder English voicebank fulfills a dream I’ve had since I was like 15, so I’ve just been fucking bouncing off the walls with excitement since yesterday when the voicebank finished training and just worked. It can still sound a lot better, but I’m happy haha. Also happy Easter, I forgot it’s Easter today. Bye bye