Discontinued Voicebanks
Hello Everyone! Effective immediately, all DiffSinger and ENUNU voicebanks have been discontinued.
If you have already downloaded any of these voicebanks, I will not stop you from using them, but please do not redistribute.
Please be aware that redistribution has always been a license violation and New York State has strict laws around the creation and distribution of digital replicas.
Our standard voicebanks are not being discontinued, and we have drastically lowered the license cost for all voicebanks aside from Zra who didn’t have an AI voicebank to begin with. I’m going to keep the CoeFont voicebanks up for now but that’s cause I had to pay to make them in the first place and nobody usin em anyways.
Why?
The primary reason is that I have grown uncomfortable with the idea of distributing an AI of my voice. Several years ago, I was briefly given the opportunity to develop for an AI engine, but it did not pan out. For some time, I felt like I had some kind of obligation to release AI models of the Z Vocal Project characters to keep up with the trends incase this opportunity ever came up again.
I was always uncomfortable with an AI of my voice floating around on the internet (see Dunder’s license originally costing $30 because it was gatekeeping the DIFF-SVC model), but originally I was willing to ignore those feelings for the sake of “advancement”. However, the more I learn about the environmental cost associated with training these voicebanks, and the ethical concerns around how the bulk of the data they use as a base was sourced, the more I have desired to discontinue all of the ones distributed here. Full discontinuation is something I’ve been teetering on doing since about August of last year, and have only finally been able to bring myself to do now.
These are personal morals and I don’t care about what others choose to do with their voices. I still use SV and V6 so this isn’t a rant against AI vocal synths or anything. Most of the commercial engines exclusively run on CPU so there’s not an extreme environmental concern with those either.
Okay? Who GAF?
I don’t think people actually check this website and I don’t really get a chance to talk about these kinds of things idk.
I also just kind of want to put an explicit notice out that I don’t want my voice trained on AI. The way some people in this community felt entitled to train AI models on people’s voices when Diff-SVC/RVC first came out, mixed with the fact there were no protections on digital replicas, kinda scared me. Its actually the reason why Dunder has a commercial license in the first place; So that there’s clear and concise messaging on what can be done with my voice.
What’s Next?
Z Vocal Project is sort of on hold right now while I work on my other project. It is not over, just not a priority. A Dunder C+V voicebank was recorded last month (along with one for a character in this other project) and I’ll probably try and get it released for June? I still have to label Solstice as well. I was a bit too optimistic in the anniversary post, my master’s program suddenly ramped up in pace and since the other project wasn’t completed in December I just fully do not have time to do Z Vocal Project right now.
I think I briefly mentioned the concat engine for use in game engines I was working on in a past article, that’s also on hold because the other project took longer to expected to release. Unless I genuinely end up thinking its something people would want to use it probably wont be a commercial thing tho.
That’s all for now, Bye Bii~