I have decided to publish a selection of my own SVS dataset for parallel training. You can use this data to help create your own DiffSinger model (or any other suitable SVS software). There has been a severe lack of easily available English female data, which motivated me to release the corpus.

I’m planning to make a proper dataset page someday, as part of the Project AIdol section.

I hope you enjoy! And please read the readme on the GitHub before using, thank you in advance!

Download on GitHub

~ Lotte V

Hello all!

I’m happy to announce that Hanami’s DiffSinger voicebank has finally gotten a full official release! As you can see, I ended up changing the name of the voicebank in order to match Project AIdol (see logo at the top of the post).

There’s some general information about the voicebank in the video description; I’m not gonna add all of that here, since that would make the post way too long. But what I will say is that she supports many different languages, has 3 different vocal modes, auto-pitch, and a custom vocoder.

I’m planning to create a dedicated DiffSinger section to the website, as well as some general information about Project AIdol. It’s probably gonna take a bit before I can do that though, but please be patient! It will happen, I promise.

You can download the voicebank here.

I hope you enjoy using her!

~ Lotte V

A new beta for Hanami’s DiffSinger model is out. Hopefully LEONA will follow soon.

This beta was trained with reflow and has 3 vocal modes instead of just 1. And again, still in beta, so apologies if there are still weird quirks around! Please let me know if you encounter anything strange.

Download link on GitHub

~ Lotte V

They’re very unfinished but enjoy them regardless. The only ones that are missing are either a) experimental, or b) not something that’d be allowed on GitHub.

My other UTAUs will come at a later date.

Have fun. I’m going to bed now

~ Lotte V

Yes, you’ve read that right! Hanami and LEONA’s DiffSinger models are now in beta. I won’t be adding them to their respective character pages yet due to their beta status, but I still wanted to have the links here on my site. So for now I’m posting them as a blog update.

Hanami download link

LEONA download link

I recommend you read the release notes before downloading; they contain some pretty important information.

Have fun!!!

~ Lotte V

Hopefully this helps some people with their projects, and hopefully the results will be satisfactory enough.

Speaking of Diffsinger… Be on the lookout for that the coming time.

~ Lotte V

If you feel like there’s any language you’re missing, feel free to suggest one! I am prioritizing cases where someone doesn’t find any of the listed languages easy to understand.

I hope this may help some people navigate the site better.

~ Lotte V

The voicebank got released a few days ago, but I forgot to make a post… Oops!

Finally, she’s here! It took me a long time, but I finally did it. This release was originally for QUiNN’s 10th anniversary, but I drew a complete blank when it came to inspiration for quite some time when it came to her design. Even though the voicebank had been finished for some time already… Oh well. I hope you like it regardless!

Her page on this site has also been updated, although as of writing this post I still need to finish the final touches. Those will come with time.

Again, I hope you like her!

~Lotte V

I haven’t mentioned this yet on the blog, but sometime last year (2022), my Blue Yeti microphone that I’ve had since 2015 unfortunately broke down. Do not fret, however! Not long after this happened, I bought a new microphone, a Rode NT-USB to be precise. This means that I can still continue recording, and in my opinion, this microphone offers higher quality audio, which means that my future voicebanks will sound even better than they did before.

However, there’s an obvious downside to all this: I have to rerecord everything. Again.

What probably bothers me the most about this, is that several of my UTAU voicebanks had either been updated not too long before my microphone broke, or I had been working on updates that were to be released in the near future. Unfortunately, in order to have a matching quality among all voicebanks, combined with my personal perfectionism, I do feel obligated to rerecord everything again. I am currently in the process of doing so.

This also means that Joëll, who had literally just been updated before my microphone broke, will have to get updated again, relatively soon after his last update. No, I don’t like it either, but again, I feel obligated to do so. Especially since I had already been planning to record an alternate English voicebank for him, recorded with a different method (more information on this in future posts), even before my Yeti broke. But I sadly can’t do anything anymore with the in-progress Yeti recordings I did have… So I have no other option. A possibly positive aspect of this, though, is that I’m currently looking into making his MidNight (soft) voice color/append tri-pitch, so that its range will match the other two colors/appends. I’m trying to make his new English bank tri-pitch as well; his current one is a bit lacking in range.

I’m also currently working on updating Hanami, whose update… was also relatively soon after the last one, but at the very least I never actually got to finish my updates for Korean and Chinese, so those will be coming, as well. And same as with Joëll above, a new English bank will come as well, same with Spanish. I’m trying to make all these banks tri-pitch.

QUiNN‘s new Japanese bank has actually been finished for a long time now; unfortunately, I’ve been lacking inspiration for her concept art, which I find rather unfortunate. I hope I’ll get the inspiration to resolve this problem.

Annushka‘s “Japanese Plus” (aka Japanese with some extra consonant sounds) is already fully recorded, but not yet fully configured. I also still need to draw her new concept art. However, I hope to release QUiNN’s update first.

LEONA had only gotten a newly-recorded extra pitch for her Core voice, so for her I’m working on rerecording her whole bank rather than a simple “patch”. I also plan on giving her a new English bank.

Regarding DeepVocal

I have some news considering DeepVocal that might disappoint some of you: While I will not take Cecilia’s existing voicebanks down, development on her has unfortunately been stopped. This also means that her planned male counterpart, Dorian Calmato, will not be released. Cecilia’s planned English voicebank, as well as the Soft, Power, and Cute Japanese voicebanks, have also been scrapped. I’m very sorry about this!

Instead, I’m considering uploading her source files (audio files/DVCFG files/etc.) of each existing voicebank to a git repository, making her effectively open source in the process, akin to my UTAU voicebanks. The reason I’m considering this is because I’m aware that her currently released voicebanks contain some instances of sloppy configuration on my part due to a lack of proper knowledge at the time, but I also can’t be bothered to fix them due to my diminished interest in DeepVocal development. So, for those who are interested, I am willing to let users themselves reconfigure her so that she will sound better, if they so desire.

Believe me, this was not a light decision to make. Cecilia as a character concept essentially represents a sort of “what if?”-situation, had I gotten into UTAU now instead of back in 2011. The existence of DeepVocal, at the time, seemed like an opportunity to me to essentially start “from scratch” with a character I’d actually be “100% behind”. However, when Hanami’s 10th anniversary arrived, I saw the opportunity to give some of the concepts that had been wandering in my mind to her as well as a sort of “character reboot”, which meant that Cecilia, unfortunately, started serving less of a purpose. This, combined with the fact that I admittedly find DeepVocal development to be an excruciatingly tedious process, and the fact that I became a dev for the wonderful software OpenUtau, has caused me to make the decisions I did. And again, I am not taking her voicebanks down – I am more than OK with people still downloading and using her! All I’m doing is step away from her development. I am setting her free.

I hope you guys understand this decision, even if you don’t entirely agree with it. I simply have a desire to concentrate full-time on (Open)UTAU again, just like how I started. It simply is my home.

So! This is the current overview of my update plans. There are a few more things I’d like to say about these updates, including something else that I have been considering, but I’ll save that for a future post.

That’s all for now. Hope to see you around!

~Lotte V

Oh, and QUiNN’s “10th” (for lack of a better term) is also still most definitely coming. Please stay tuned for this one as well!

~Lotte V