No.21484[View All]
A new singing synthesizer by ExpressiveLabs that claims to be easy to use and compatible with UTAU voicebanks. The crowdfunding campaign starts on April 1st.
https://x.com/expressivelabs/status/2036956525223113012https://mikoto.studio/crowdfund<Can i become a voice provider?Yis...
https://expressivelabs.net/talent No.24683
>>21484I misread the OP as "Singing synthesizer by ElevenLabs" and thought the "I'm sorry but I can't fulfill this request" guys were making a vocal synth.
No.25773
Now that the dust has settled what is /leek/'s veredict on Mikoto
No.25774
>>25773I mean it's alright. I don't see myself using it over OpenUTAU and there's definitely room for improvement but overall a lot of the problems mentioned ITT are overstated and Mikoto's just an okay vsynth.
No.25956
The campaign has finished
"Project was not funded
The project didn't meet its goal.
No funds were collected from backers.
The campaign ended on 5/15/2026.
56 backers pledged €18,197
This is 51.99% of the €35,000 goal."
No.25958
>>25956What does this mean for Mikoto
No.25962
>>25958>"this isn’t the end of Mikoto Studio. We are still 100% committed to bringing our singing synthesizer to life. The campaign might not have gone our way, but the dream is very much alive. The only difference is that it’s just going to take us a little longer than we originally planned. We’ll be sure to keep you updated as the development continues."
>Right now, though, our small team is going to take a short 2 week break to rest, reflect, and recharge after the intensity of the campaign. We’ve poured everything into this, and we believe coming back with fresh energy is the best thing we can do for the project."Also this:
>"mikoto 0.5.5 is still coming before our 2 week break, stay tuned for that" No.25970
>>23817She sounds better in Mikoto Studio but it bothers me that her voice just cuts off like that
No.25978
Honestly they just should have never did a crowdfund. It was doomed to fail. Hopefully they make do with this back up plan though.
No.25994
>>25978Is it because western studio? I noticed JP crowdfunds get funded within hours if not minutes
No.25998
>>25994I don't think so, western fans have been crowdfunding stuff for a while. Mikoto's problem was that they only really marketed to overseas UTAU users. That part of the fandom isn't big enough to carry a 30k crowdfund.
No.26000
>>25998Imo it was because their thing was that it's an "easy alternative" to utau, which doesn't hold when you market to people that grew up on it, and when OpenUtau offers the same functionality, but it's free. They should have marketed to people that are new to vocal synthesis.
No.26001
>>26000Someone should make a sample based vocal-synth. I know that AI has made the demand for those lower and that UTAU exists but I don't think it'd do too badly.
No.26003
>>26001There's kinda no need for one honestly. UTAU already exists and has an infinite library of free voicebanks and with resamplers you can improve their quality. Best part is that its free.
No.26011
>>26003Yeah i agree, OpenUTAU with MoreSampler and HiFiSampler is all you really need these days. For voicebanks you can just pick a voicebank that's recorded in decent quality.
No.26020
>>23817Sounds less metallic than moresampler but the cutoff is weird
No.26057
>>25998>>26000These plus the fact lingering factors from Maghni not having come out yet as well as the fact they were not marketing to professionals properly. In the west you NEED the support of professionals.
Not to mention, none of the JP crowdfunds were necessarily for new engines either. They were for voicebanks for existing ones, so it’s not necessarily a fair comparison.
No.26058
>>26057Adding on again: this is also their very first product. They are a new company. By comparison VocaTone at least had Oliver and Yohio on Vocaloid so people had the impression of deliverability. Of course sentiment in that regard changed but during the campaign, having something delivered was better than nothing.
No.26060
>>26058They also had Big AL, Sweet Ann and RUBY/RVBY. All of which were teased...and then all we got was a RVBY demo
No.26061
>>26057>In the west you NEED the support of professionalsFWIW, Maghni hit their crowdfund goal back when all of their promos were targeting the fandom. I could see Mikoto being funded if they had more big names attached to the project.
No.26063
>>26061This and the fact that they promised AI versions of those old voicebanks which are stuck in V2-3. AFAIK V6 is no longer compatible with V1-V2 and chances are they'll drop support for V3 by the time V7 is out.
No.26092
Meika Mikoto Studio
No.26103
>>26061This is not entirely accurate. If you look a bit deeper, they had backing from universities and showed Maghni’s development off at in person events. The goal for Maghni was also less than Mikoto’s. There was more motivation after they reached the goal, but Maghni had attracted professional attention despite its appearance.
No.26105
>>26063As long as V5 is supported, v3 should be supported. I don’t think we’ll have to worry about support ending as much as we’ll have to worry about Maghni being in dev hell.
No.26106
>>26105They did kill V2 support though didn't they? Voicebanks like Big AL and Sweet Ann never got an update.
No.26110
>>26106Yes, but they killed it off halfway through V4. If V3 support was gonna end, they would’ve done it years ago. Not to mention, they re-added and added features for V3-V5 banks in V6 like XSY and direct pitch. We’ll see more V3 banks discontinue before they end support.
No.29843
>>26106Thank god for tunelab
No.30091
So what happened to this project?
No.30095
>>26106It's really disappointing that if you own anything newer than V4 you basically can't use them and have to resort to either buying an older version of Vocaloid or
Tunelab pirate plugins.
No.30105
>>30091On hiatus apparently, idk if their dicksword has anything more specific
No.30226
>>30095You can replace the files of a V6 compatible bank and it'll work but at that point I'd just use Tunelab
No.30227
>>30105To be honest, i don't know why they thought making a paid version of UTAU was going to work
I want western vocal synths to succeed but so far only Maghni brings something new to the table and even that failed
No.30240
>>30227I wouldn't say it has failed YET but its future isn't looking too bright
No.30242
>>30240They said they're still gonna finish it at least just slower
No.32460
Not to necro but I just wanna complain somewhere. Approached them, had a meeting, and was looking forward to developing a line of voices as their tech seemed promising. I had a plan and character documents etc ready, but after the initial talks (which I believe went well!), I got totally ghosted. After multiple attempts to get some clarity on the sudden lack of communication, I still got nothing. It's just... lame. Really lame. "Wtf" really, you don't just do that. You could have said anything, or even told me you were reconsidering, but just complete silence? Unprofessional af..
No.32462
>>32460Why is it that all the vocalsynth companies here seem really unprofessional? Maghni's the only one I kinda trusted due to their association with Vocatone and look how that turned out.
No.32464
>>32462> vocalsynth companiesMakes me wonder if the server they all congregate in is an echochamber of sorts. I'm not optimistic for newcomers in the commercial scene (if you can even call it that).
Also, Mikoto's AI is basically Diffsinger (they even use the diffsinger format). Also, their NEST engine is similar to Vocaloid's concat engine, but I'm preferential to Vocaloid. If one just wanted to sell their own line of voices, professionally recorded and without insane licensing fees (Mikoto's was fair, I'll give them that), just stick to Diffsinger and UTAU. I know some people get flack for commercial voices, but Mikoto at present doesn't offer anything to the experienced vocal synth user.
TBH, Mikoto as a software suite is marketed towards those who aren't experienced with vocal synths but still want to get into it.
No.32465
>>32464>Also, Mikoto's AI is basically Diffsinger (they even use the diffsinger format).??
No.32466
>32465
Partial mistake here, they use the same methods as diffsinger for *data annotation*. As for synthesis, I can't make definitive comments. Sorry about that.
No.32486
>>32481monkey's paw strikes again
No.32506
>>32464There are outliers, but most of the companies that have a lot of these problems tend to be Western ones.
The Eastern side of things has its hiccups, but those happen much less by comparison. I swear almost every Western company has some sort of issues going on behind the scenes.
At the very least, I hope these recent failures actually become learning experiences.
No.32507
>>32506What western vocal synth company hasn't fumbled though?
>inb4 Eclipsed SoundsReferring to companies that have made their own vocal synth engines here
No.32515
>>32466I think ENUNU uses the same method so it's not a DiffSinger thing. I'd go as far as to say that most AI vocal synth labeling probably follows a similar process if not the exact one.
No.32564
>>32507Pompeu Fabra University
No.32570
>>32462The broader issue is that most businesspeople in the west don't have a good grasp of otaku culture, nor do they see it as worth the effort. Meanwhile, the people who really want to make vocal synths are all fresh college grad programmers and artists who haven't been in the corporate world. There's a lot of basic practical knowledge that isn't being taught here because the overlap is so small.
No.32582
>>32570You hit the nail on the head
No.32594
>>21485If they ever release a commercial voicebank it's not gonna be on a western engine. Just saying.