To start off, Iâm not in the position to complain about the banks Iâm asked to review. I wonât complain for that matter either. The only exception I make is when a voicebank messes with my system, and even then I wouldnât completely call it complaining.
The voicebank we have here is
Kyokusei Jiseki
Itâs a CV style male bank with about 132 samples whose first web presence I could track down to 05.05.2012.
The bank has a young, boyish tone to it while the intensity is pretty strong. Not Ritsu KIRE strong, regular strong. When used the voice sounds a bit muffled due to the recording quality, but that isnât necessarily a bad thing.
Speaking of the quality. The bank seems to have been recorded with a built-in microphone or a cheap headset, thus being of rather low quality with quite some noise. Again, a low quality doesnât automatically make it a âbadâ voicebank. The samples provide roughly half a second long actual audio to work with.
When looking at the oto.ini, I noticed that the samples for ã¡ã, ã¡ã
and ãã
were missing, as well as the j- and y-combos being the only ones to have -e-samples. The pronunciation is spot-on and an accent only slightly noticeable on the âo-samples. Iâd like to note that the bank provided r-samples as well as l-samples; the r-samples had the proper pronunciation (which I am biased to like) and thus itâs nice to also have l-samples in case you want to, I donât know, go through the pain of doing engrish or if you need lalaing.
The oto.ini-configuration itself was alright. Again, first thing jumping into my face was the near-total lack of overlap values. Only one sample actually used it.
Quick break: Overlap defines how the previous note blends into the one youâre configuring. Giving it a positive value will let the previous note blend into this one (commonly used for what I like to call soft or stretchable sounds, i.e. s-, m-, n-), whilst a negative value will shorten the previous note, resulting in a âgapâ (think the pause when saying k-, t-, d-, the hard sounds).
So by now youâre probably like âShadow, thatâs nice and all, but what the **** were those three first sentences for in that case?â
This voicebank only works with one sampler.
This voicebank only works with one ****ing sampler.
Iâd been told beforehand that one sampler didnât work with this bank. When I wanted to test how itâd work with a sampler apart from resamplerâ¦
First the sampler crashed.
And then UTAU followed.
Iâve been working with UTAU for more than three years now and Iâve hardly seen something like this happen. Thereâs about five or six samplers in the folder and every single one of them apart from resampler refused to work here.
Well, I thought to myself, canât be helped, Iâll just work with resampler in that case and look for a solution later on. Itâs a damper on the available usage freedom to only have the voicebank work with one sampler, but not something to classify something as a bad bank.
From time to time while working with this bank I switched to another window of UTAU to work on my own stuff. After a while I frowned: Nearly all of my voicebanks are kept on my server in the cellar, so by now I was used to some banks generating sound more slowly than the two or three banks I keep on my hard drive, but this time around it was even slower than usual. After about five minutes of generating it played and I switched to TIPS. Again: really long generation time. Shrugging, I didnât plan on rendering the full wav file of this today anyways, I closed that window and went back to the internet, where a nice video that piqued my interest crossed my path. While I watched I got an IM from a friend and thatâs when all hell broke loose. I tried pausing said youtube video and an immense lag started before it finally properly paused. There were some more occurrences, but long story short, somehow the usage of this voicebank had completely ****ed with my CPU.
âBut Shadow, couldnât that be because of the heaps of programs you had opened?â
No. As I write this I have opened about the triple amount of stuff than yesterday when all that happened. Using the voicebank just seems to have put a damper on the CPU until I shut down and restarted my system.
Nearly as soon as I had noticed all of this I got my lazy ass to look for what was causing all this shit. I compared the wav-files to others from other banks I own. They were the same except for one tiny tidbit: These files here had two channels. Again I frowned and went to ask another friend, this one having worked with UTAU for about two year himself now, if this had ever been a problem with UTAU. He said that no, normally not, and that his UTAUloidâs first bank had also had samples with two channels. Afterwards I meddled around with one of the wav-files and got to export it as a single channel file. After the five minutes UTAU took to boot up and load the voicebank I gave it a test.
Resampler: works.
TIPS: works.
Bkh01: works.
So now I am not sure if I can pinpoint it to being due to the samplers being racists nowadays, UTAU being racist or if itâs really due to the voicebank.
Conclusion:
A voicebank with potential which clearly needs rerecording to fix that huge channel-****ing. Iâd only recommend it to users whoâd go through the effort of converting the 132 samples to single-channel and meddle a bit with the oto.ini. The others? Wait for an update. Once that get fixed, youâll surely like working with it.