Real-time voice conversion software aimed at VTubers called "Voidol"
- Thread starter uncreepy
- Start date
I've used Rice-chan with English and Japanese and she worked to the same level of quality as all of the other voices I have. I don't know if for iPhone the download of her bank is region blocked or something? (I don't have iPhone, so I can't check.)
Voidol updated to 1.3 (both Mac and Windows version). Just start up the app to be prompted to download the update.
The update has improved voice quality of voice models for the following characters:
Zunko, Iroha, Crimmzoh, Minato, Cutie Angel Peroro,
Yoshida-kun, Miranda, Tsukasa, Tsubasa, and Rice-chan
I downloaded the update, nothing changed with the interface. My anti-virus scanned it a few times, too. Anyway, I thought the sound quality didn't seem SUPER different, some patterns didn't seem to have the ultra choppy explosive noises (only said a sentence per model, so I need to test more). I think they added more patterns for Minato. I will take a better look at it and post an update in here tomorrow when I have time.
(I hope this means the VST is coming relatively soon.)
The update has improved voice quality of voice models for the following characters:
Zunko, Iroha, Crimmzoh, Minato, Cutie Angel Peroro,
Yoshida-kun, Miranda, Tsukasa, Tsubasa, and Rice-chan
I downloaded the update, nothing changed with the interface. My anti-virus scanned it a few times, too. Anyway, I thought the sound quality didn't seem SUPER different, some patterns didn't seem to have the ultra choppy explosive noises (only said a sentence per model, so I need to test more). I think they added more patterns for Minato. I will take a better look at it and post an update in here tomorrow when I have time.
(I hope this means the VST is coming relatively soon.)
I have Iroha, Minato, Crimmzoh, and Rice-chan. I checked and all of them have the same number of voice models as before. (Last time I posted, I thought the number had changed, but I was wrong.)
I never posted it in here, but I did record myself saying the same sentence with each character/each model in this tweet thread:
So today, I have recorded a new tweet thread with new recordings and I think it has improved quite a bit!
There is a problem with the audio that is only noticeable with headphones (each time I talk, there is a loud burst of "air" that kind of hurts your ears, I tried changing the settings for Voidol and my mic but nothing made it go away). It sounds actually pretty fine on normal speakers, though.
The mechanical gargle sound from before seems to be for the most part gone. (The results still aren't perfect, but this is significantly better than before in my opinion.)
I also want to test out speaking in Japanese/singing, but not at this moment.
I never posted it in here, but I did record myself saying the same sentence with each character/each model in this tweet thread:
So today, I have recorded a new tweet thread with new recordings and I think it has improved quite a bit!
There is a problem with the audio that is only noticeable with headphones (each time I talk, there is a loud burst of "air" that kind of hurts your ears, I tried changing the settings for Voidol and my mic but nothing made it go away). It sounds actually pretty fine on normal speakers, though.
The mechanical gargle sound from before seems to be for the most part gone. (The results still aren't perfect, but this is significantly better than before in my opinion.)
I also want to test out speaking in Japanese/singing, but not at this moment.
Sounds like the quality has improved a LOT since previous versions. I'm still holding out for that VST version, but I'm glad to see the software is improving.
Crimson Technology announced that the VST version of Voidol is planned to be released at the end of this month!
The wording is a little vague, but it says:
They say the image could change since they're still finishing it up. But it pretty much looks the same as the English version they teased in January:
In February, we had found out that they were working on singing voices for Otomiya Iroha and ANDROID GIRL (still don't know who that is), and talk voices for Tohoku Zunko and Otomiya Iroha. It's been a long time since then, so hopefully they updated other characters for their talk/song voices.
The wording is a little vague, but it says:
They say the image could change since they're still finishing it up. But it pretty much looks the same as the English version they teased in January:
In February, we had found out that they were working on singing voices for Otomiya Iroha and ANDROID GIRL (still don't know who that is), and talk voices for Tohoku Zunko and Otomiya Iroha. It's been a long time since then, so hopefully they updated other characters for their talk/song voices.
@Maria No, they are definitely not using Voidol. Voidol sounds obviously mechanical/low quality, here are a couple examples of Voidol:
Voidol is a real-time deep-learning-based voice changer. A lot of VTubers actually use other voice changers that only change formants/pitch. The most popular voice changer is actually called Koigoe. Here is the link to another thread where I translated information about using Koigoe/about its interface:
vocaverse.network
(Please see posts #3 for general voice changer information and especially #5 and #6 for instructions on Koigoe. I actually have to translate more information for that thread, but I have some personal things going on and can't do it at the moment. Hopefully in a few days.)
Voidol is a real-time deep-learning-based voice changer. A lot of VTubers actually use other voice changers that only change formants/pitch. The most popular voice changer is actually called Koigoe. Here is the link to another thread where I translated information about using Koigoe/about its interface:
On voice (technical aspect)
Good day. Could someone explain me please how vtubers change theirs voices? Does it need additional devices? I tried some software like voidol, voicemode, etc., but couldn't achieve my voice to be even approximately like theirs. Moreover some background noise always comes with my speech while I...
vocaverse.network
(Please see posts #3 for general voice changer information and especially #5 and #6 for instructions on Koigoe. I actually have to translate more information for that thread, but I have some personal things going on and can't do it at the moment. Hopefully in a few days.)
I don't know what happened, but Crimson Technology hasn't actually written an announcement about this and you can't actually buy Yukari or the VST on Amazon yet (they're there, just not purchasable yet). You can buy them on Rakuten, it looks like, though.
Voidol VST = $83
Yukari = $46
I was hoping the Voidol VST wouldn't be this expensive, I thought it would be like $45ish at most, but twice that amount. Arg. I don't really want to buy it on Rakuten, either, so I might wait for the Amazon listing to "wake up".
Voidol VST = $83
Yukari = $46
I was hoping the Voidol VST wouldn't be this expensive, I thought it would be like $45ish at most, but twice that amount. Arg. I don't really want to buy it on Rakuten, either, so I might wait for the Amazon listing to "wake up".
Though there isn't an official announcement, this page did update to mention the VST:
There are actually 2 new voices on the voice model page
They are an idol duo called simpαtix (シンパティクシュ SHINPATIKUSHU). The members are Tsukina Takai and Terashima Yufu. They probably have them because of Voidol VST's singing models.
There are actually 2 new voices on the voice model page
They are an idol duo called simpαtix (シンパティクシュ SHINPATIKUSHU). The members are Tsukina Takai and Terashima Yufu. They probably have them because of Voidol VST's singing models.
Last edited:
1: VOICEROID Yukari
2. Hiho's own voice
3. Hiho's voice being converted with Voidol
4. The AI voice changer Hiho made on his own with Yukari (I don't know if anyone remembers from VocaloidOtaku, but he was making his own real-time voice converter using Yukari's Voiceroid voice, he kept developing it over the years and he is actually now an engineer for Seiren Voice)
Apparently now there are 6 voice models for men instead of only 2 (I am not sure if it's only for Yukari) because of the VOCALOMAKETS guys contributing to it.
Still waiting to be able to BUY Voidol VST. Ugg...
They accidentally leaked possible upcoming voices again during a demo for DTM Station (this is the video):
Pepper got leaked during a DTM Station live stream on June 4, 2019. ANDROID GIRL got shown in a screenshot in a DTM Station article from January 17, 2020. The two new voices are:
しろたん Shiro-tan (spelled Sirotan by Creative Yoko), he is a cute white seal (shiro = white, "-tan" = cute version of -"chan" honorific).
しんちゃん = Shin-chan. I don't know if it's THE Crayon Shin-chan or not.
From the interview with DTM Station:
It sounds like if you purchase additional voice banks and only planned on buying the VST version of Voidol ($83), you unfortunately also have to buy the standalone version ($20) in order to actually "activate" those new voice banks in the VST version.
I downloaded the trial of voidol the other day, but it keeps sounding really staticy. I am just using a headset, and a virtual cable to output the sound so I can record it in audacity. Is there a common way to make the sound clearer? I don't have quite as much static when I use other voice changing software.
Voidol literally just sounds that bad. It's not a normal voice changer that just affects pitch/formants, this one has deep learning as the gimmick to turn your voice into a specific character's rather than just editing your voice to be higher/lower. The last update improved it a little, but it's still far from convincing technology. Please check out some clips by other users to compare the level of quality you can expect:
The new singing models seems to be a lot more convincing with the release of the VST, but I haven't seen anyone demoing it other than this official one because it's still not available on Amazon/there are troubles with Rakuten:
When I used the demo and guided other people through the process, the trial only lasted for 5 minutes, so I don't know if you will be able to try it, but here are some tips: 1) Try out each voice model. For example, even though I am a girl, sometimes using the male patterns gives better results depending on the character. 2) Practice speaking fast/slow and even try voice acting (ex: if I want to use Minato, I should talk in as low as possible voice as I can so the "leap" between female to male conversion isn't as huge). 3) Crimson Technology recommends a cardioid mic.
The new singing models seems to be a lot more convincing with the release of the VST, but I haven't seen anyone demoing it other than this official one because it's still not available on Amazon/there are troubles with Rakuten:
When I used the demo and guided other people through the process, the trial only lasted for 5 minutes, so I don't know if you will be able to try it, but here are some tips: 1) Try out each voice model. For example, even though I am a girl, sometimes using the male patterns gives better results depending on the character. 2) Practice speaking fast/slow and even try voice acting (ex: if I want to use Minato, I should talk in as low as possible voice as I can so the "leap" between female to male conversion isn't as huge). 3) Crimson Technology recommends a cardioid mic.
The 3 new voice models for Voidol are now working correctly on Amazon Japan (Yuzuki Yukari & the two simpαtix girls), but the VST is still not available. Crimson Technology said on Twitter that they are hoping it will be available soon.