2022 Art Update

Thanksgiving Thanksgiving Thanksgiving.

After practicing a new shading style for the first half of this month by way of Zealous Villain Project, I’ve decided to update the art for all the Z Vocal Project characters aside from Dunder.

Some characters are still just in their Z Vocal Project outfits, and that is because I don’t want to commit to anything right now. Fuuka, Dawn, and Kiyana have had their idol outfits improved upon a bit, these are not their finalized designs, as whoever they get handed off to for commissioned portraits will be asked to fix them up a bit more.

Kemmy is now properly represented with art for her “Neutral”, “M” and “F”, outfits, as opposed to her school outfit art, which was just a toned down version of her “F” outfit. I’m gonna redesign her sword eventually, my first attempt wasn’t the best and I want her to have just a straight-up di-sword so god knows when I’ll figure that out.

Herculo’s new art was drawn at an angle for the sake of translating what he’s supposed to be better than the head-on art. Like Kemmy’s art, this was meant to be out last year shortly after the relaunch, and I apologize for the delay, I hope it’s a bit more clear now what their concepts are now.

A Hive Has Been Created

We are on Hive under the name ZVocalProject. and Dunder under the name Dunder. If Twitter dies we will move there, but have no plans to move otherwise.

Dunder_SVC in development

I say in development but really it’s just a matter of recording a lot more data. The following sample was trained on all of what I currently have recorded for Dunder AI (about 1 hour). This is mostly soft mid-range data as my sleep schedule’s fucked and I can’t be belting at 3 am.


Thought parts of why this sounds so bad is a low amount of data, a lot of it is due to half of the input data being me keeping my voice down. This is was trained for like 80k steps so I know it’s not a training issue, I might end up just not using this data in the future now that I know how much it’s actually going to affect the end product, but I will still try mixing it in with the louder and better data going forward.

The following is an actual singing sample, and what I’m probably aiming for for the tone of this voicebank.

Actual Singing

As you can probably tell, I cannot sing and my microphone ain’t the best. Dunder_SVC is more of a test than something I actually intend to release. I think at best it could end up being useful for backup vocals, but really it’s so if I end up releasing an SVC bank of one of the other characters I know what I’m doing. I’ve also seen people using UTAU renders to train banks with some success, so I may make a version of SVC based on that as well. All that said, here is the official art for Dunder_SVC:

Dunder_SVC’s key art is a low poly model made by Mazata. They made this for me in February of 2020 and I’ve been thinking of a practical use for it for some time now. The low-poly model was chosen due to its ties to a specific era in gaming just before the split between realism and stylization. Diff-SVC is both the continuation of dreams of being able to breath lives into our characters, and the anti-thesis of creating a singing synthesis voicebank in the first place. You may say “This argument makes more sense with AI vs Standard banks” and to that I say, you’re right, but I want the ENUNU bank to have normal art, so this whole metaphor is getting placed on SVC.

(I tried porting the model to MMD but I couldn’t get the IKs right, but I did get it working in clip studio paint.)

Clip Studio Paint shitpost

Happy thanksgiving and all that. next post will probably be in December.
Edit: Dunder_Intercalaris+ is now in development