Real-time voice conversion software aimed at VTubers called "Voidol"

[Nyanami]

Thanks for your help,

I tried your method with Audacity but as I didn't have the same settings as you (Sound Blaster & Stereo Mix) I came up with something else.
I used a virtual audio cable you can find and download here : VB-Audio Virtual Apps
I think you will have to restart your computer after downloading it.
With this you have 2 things : CABLE Input and CABLE Output.

By applying the output of Voidol as CABLE Input and the microphone of discord as CABLE Output, it works and does the voice conversion.
Moreover, you don't hear yourself talking, as you mentioned, it is pretty hard to speak while hearing yourself.

I don't have a real mic though, only a microphone on my headphone. Do you think there's a lot of difference between the two ?

 

[uncreepy]

I don't have a real mic though, only a microphone on my headphone. Do you think there's a lot of difference between the two ?

Sorry for replying really late, I somehow missed the notification. You are able to use any type of mic, it's just that Crimson Technology recommended that Voidol works best with a mic + audio box.

When I get time, I will definitely try out your method of using a virtual audio cable!


As a side note, Crimson Technology announced they are developing a VST version of Voidol (thank goodness, the current way to record is a bit inconvenient). It's a Steinberg VST. They will present it at The 2020 NAMM Show in America on the 16th-19th.


I wonder if people who bought Voidol's standalone version will get the VST for free?


Also, they revealed in the article that CRIMMZOH is supposed to be a young boy.

 

[frankensalad]

Once that VST version gets released, I'll probably buy it. I know that Cherry Pie sounds better (at least from the pre-release stuff we've heard, which I'm taking with a grain of salt) but the main reason I'm interested in this software is for providing voice overs for youtube videos without using my actual voice, and I don't want to use voices that are as recognizable as the Cryptonloids.

 

[Nyanami]

What will be the difference between the VST and the normal version ?

 

[frankensalad]

VSTs can be loaded as either instruments or effects in a digital audio workstation. The VST version of Voidol would most likely be of the effect kind, meaning you would be able to make a recording of your voice, clean it up in your DAW, and THEN run it through Voidol for more appealing results. It would also be a lot easier to capture the results for a youtube video or something. The current normal version was designed with streaming in mind, so it was made as a standalone software that takes audio directly from your mic and converts it in real time so you can stream with an altered voice. Which is cool, but the results aren't as nice as they could be if you had the option to clean the vocals up before hand and record multiple takes, which would be preferable for youtube videos.

 

[Nyanami]

Ah I see, so the VST version is more for editing, like recording different sentences for youtube videos and the normal version is more for a live conversion.

 

[uncreepy]

Here's the official music video for TOKYO "Mousou" 2020 feat. Otomiya Iroha By LUDWIG 3rd
(I mentioned it a while back.)


Watch out, Hatsune Miku!

By the way, I made a tweet thread that contains demonstrations of each conversion model for each free character (click the tweet for the other 5 tweets):

 

[Leon]

I saw your demos! It's really interesting to hear. Some of the conversions sounded decent and some were just,, well. Hrm.

 

[Prism]

Any news on vst version

 

[uncreepy]

Nothing official on Voidol's website, but I did find an... er... interesting video of the booth at NAMM 2020 demonstrating the VST (this is literally the only report on it being at NAMM that I could find). They demonstrate its capabilities through karaoke (which, while the video is kind of awkward, is actually is quite amazing because you can sing and have your voice recorded/converted while the song plays out loud in real time).

Timestamped for the good part:


But, based on the interfacee of this VST, I feel like it is not ready for commercial sale yet (it says "Demo" in the top right corner). The voice being used is just called "male44100" and there are no images of the voicebank being used like the phone/PC versions have.

 

[uncreepy]

Today is a sad day in vocal synth land... Voidol got reduced to one lowly star for its Amazon ratings.

But I believe it has a chance to redeem itself with the VST version!
...Maybe!
... Sort of!
...Okay, probably not, but I'm rooting for it.

Still no reviews on any other add-on voices other than Zunko/Kiritan, which some poor sap bought both at once and was surprised by the bad quality and left a scathing review... Poor soul should have bought one at a time. RIP, dude's wallet.

I also noticed that they updated the product page to advertise that companies can get the SDK in order to make their own Voidol voice banks.

 

[uncreepy]

I found a DTM Station article detailing a bit about the VST version of Voidol. I will summarize juicy parts.

The full name of the plug-in is "Voidol -Powered by R.C. voice VST Plugin version", but the article decides to refer to it as Voidol VST after mentioning it. (Seriously, such a long, dumb name!)
The version shown at NAMM was a prototype.

The VST version will have a price and there isn't a release date scheduled (they'd better not, they were late 3 times last time, better to announce it closer to a date they're confident with or just surprise us).
^ For people who have already bought Standalone Voidol or any add-on voices, they said that you will be able to switch between both versions. However, the song voice model is going to be developed as a new product in the future that you should expect to pay for.
It is not a VST instrument, it is VST effect (remember how Crypton says they have "effectors"?)

This VST uses a different model compared to the current Voidol. The current Voidol's model is based on talking, the VST version will be modeled around singing. The TOKYO "Mousou" 2020 song uses this new singing model (which is in alpha). They say the current Otomiya Iroha that people own in the standalone version can't sing as well as in this demo video, so you'll have to wait for the VST to make her sing that nicely.
There are currently 4 prototype voices: Otomiya Iroha (talk model), Otomiya Iroha TOKYO 2020 (song model), ANDROID GIRL (song model), and Tohoku Zunko (talk model). Talk voices can be used for songs, but they won't sound as nice as song model versions.
Great, no male versions yet... : { I have no clue who Android Girl is supposed to be, must be a new character because none of the current add-on voices are androids (whoever they are, they must be a big fan of DECO*27 lol).




The author of the article says that it felt similar to using VocaListener. It converts in real-time with a tiny amount of latency, but more latency compared to the standalone version.

Parameters included in the VST are Noise Gate (if there is noise on the original track, it will be tweaked to ignore that noise), GV Coefficient (changes the degree of voice conversion), OFFSET GAIN (adjust loudness of the converted voice), LIMITER (part of the last step, recommended to leave on).

 

[uncreepy]

Wow. I went to investigate the Crimson Technology website and realized I've been calling Gate Jobs the wrong name since AUGUST 2019. I kept calling him Get Jobs (and even called him Good Jobs once). So I edited posts #24, 51, 64, 68, 72, and 76 to be Gate Jobs. I realized it because it says how to read his name in this URL:

ゲート・ジョブス(CV:AIJ)

AIJさんがCVを務める「ゲート・ジョブス」の声になりきれるボイスモデルです。ボイスチェンジャーVoidolとともにお使いください。

crimsontech.jp

I swear I can speak Japanese. orz I guess I just read the name too fast before, assumed it was Get, and never double checked. ゲット = get　ゲート = gate. (This is extremely funny yet highly embarrassing to me...)

 

[frankensalad]

To be fair, English rendered in Katakana is always a wild ride, and Gate Jobs isn't exactly a ... normal name.

 

[uncreepy]

We've been waiting for Voidol news, but I saw Crimson Technology shared a mildly interesting bit of news just now: Iroha is the mascot for AImelo, "the world's first AI engine for Apple Watch" that composes a song automatically based on your heartbeat data depending on the type of feeling you want it to have. The original version is brAINMelody, this is just a mini version.

Thought it was cute/interesting they are using her. Hopefully her idol "career" extends further than just that Tokyo 2020 song.



The news from Crimson Technology lately has just been about the virus (that they are working remotely). They've had basically the same virus announcement a whopping 7 times since March, most recent being April 12th. So I am actually surprised yet glad they announced this brAInMelody thing, it shows they are still working on products. The latest announcement says they will still be working remotely until May 6th and will decide to extend it or not then. I wonder if we should not get our hopes up about Voidol VST anytime soon until after they return to the physical work location? It's not like they're going to sell a physical version of Voidol, though, so maybe they could release it at any time.

 

[Trevor]

Given the fact that Crimson Technology is just a passion project, I think we can expect to see continued development. brAInMelody is just the practical application/proof-of-concept of the creator's research and thesis statements. I would be more than happy to pull up the academic papers if anyone is interested, but I highly doubt any of their projects are profit-motivated at this current moment. The creators are probably aiming to sell the software's algorithms to an established company. Cimson's development team is more-likely-than-not based in a small closet inside a university. Working from home shouldn't impact software development too much. It sorta exist just to be put on a resume.

(I jut looked it up... it really is just a closet at a university)

 

[DefiantKitsune]

^ I rearranged the list to be in numerical order. I dunno what I was thinking before.

I checked the Amazon Japan sales rankings themselves for each character. This is based on how many people actually purchased each add-on voice and it is updated every hour in relation to the categories the product is in. Voidol and its add-on voices are in the Software category.

Voidol trial version: #7
Voidol full version: #421
Queen Shuffle: #603 <-- What the heck?? HOW?! How is she more popular than the Tohokus? Are there that many "Queen Shuffle-sama" spouting Japanese LARPers?

Kiritan: #1,202
Zunko: #1,684
Tsubasa: #2,126
Ichiru: #2,171

I know I'm like 4 months late on this but I'd guess most Tohoku fans go for the more expensive but better Voiceroid/Vocaloid/UTAU banks (which whoever QS is doesn't have obvs)

 

[uncreepy]

Voidol made a press release that the iPhone version of Voidol will be updated tomorrow (the 8th) that includes all the new add-on voices.

It also says that Yone-chan/Rice-chan is supposed to be Chinese language exclusive (for phones apparently, because I got her for free with the Voidol for Windows update a long time ago). They also revealed that her voice provider is Sone Wakana (Baidu article here).
They also updated the add-on voice page to include Rice-chan, so you can hear her demos here:

ヨネちゃん（CV 染音若蔡）

crimsontech.jp

Her outfit in the big picture is different than the smaller ones, the smaller ones match what my Voidol has her wearing. It says she attracts customers for Midifan. This is Rice-chan's Chinese name: 米饭酱 and I found some pictures of her as the mascot here and here.


I found Sone Wakana's YouTube and she sings in English and Chinese, here's one song:

It also seems like she is a voice actress for games and anime like RWBY.


The add-on voice page also includes Iroha Otomiya now, her demos are here:

音宮いろは（CV遠野まゆ）

crimsontech.jp

It says that Iroha was born recently and is curious and pure and wants to know more about the world.

 

[frankensalad]

I'm glad to see they're adding more voices to the base package. I'm still waiting for the vst version.

 

[uncreepy]

I'm hoping this is them just getting all the versions of Voidol in order so they can prepare for the VST release. I'm getting excited, hopefully we won't be waiting much longer!