i want more web/mobile based synths. The only tunable ones that exist are saki web, voca mobile and ace. And even so, Saki does not work on mobile, voca mobile is hard to get and ace isn't fun to tune on mobile
Unpopular Opinions
- Thread starter TheStarPalace
- Start date
Oh, I do agree that the noisiness has diminished overtime. It's just something that continues to bug me about CeVIO and stops me loving it as much as Vocaloid and SynthV.
I'd be down for this if more companies offered support for browsers other than Chrome. It feels like we're back to the bad old days of Internet Explorer dominating website coding and design, where some sites would treat you like a proper weirdo for using an alternative browser.
Honestly Cevio seems to improve in terms of engine noise lately. Sasara's last update was a major improvement and newer releases tend to have less than older ones
So it's been about a month since English-speaking vocal synth fans blew up on Mitchie M for posting a silly AI-generated image of Miku playing a saxotrumpet on Twitter, and now English-speaking vocal synth fans are going crazy over Diff-SVC, which is based on the exact same technology as NovelAI (diffusion model machine learning). I'd just like to ask: what exactly makes one of these techniques inherently unethical and bad for human artists everywhere, and what makes the other one just a neat new thing to mess around with?
For the record, my opinion on both of these tools is that they're ultimately morally neutral by themselves, and it's entirely on the user whether they're utilized for good (shitposting, supplementing "traditional" art, or giving unskilled artists/non-artists a chance to "create" something are all perfectly acceptable use cases to me) or for bad (art theft, stealing job opportunities from human artists, honestly you've all probably heard the arguments against it at this point). And honestly, for all the bad press "AI art" has gotten about its supposed potential harms to society, I feel like giving randos the ability to clone people's voices with relatively little effort is way more risky on its own. Art theft is bad and hurting artists' livelihoods is horrible, but there are so many awful things one could do with voice cloning that I'm struggling to narrow it down to one or two examples for this sentence. Imagine someone using your voice for like, hate speech or something and that permanently affecting your career or social life.
For the record, my opinion on both of these tools is that they're ultimately morally neutral by themselves, and it's entirely on the user whether they're utilized for good (shitposting, supplementing "traditional" art, or giving unskilled artists/non-artists a chance to "create" something are all perfectly acceptable use cases to me) or for bad (art theft, stealing job opportunities from human artists, honestly you've all probably heard the arguments against it at this point). And honestly, for all the bad press "AI art" has gotten about its supposed potential harms to society, I feel like giving randos the ability to clone people's voices with relatively little effort is way more risky on its own. Art theft is bad and hurting artists' livelihoods is horrible, but there are so many awful things one could do with voice cloning that I'm struggling to narrow it down to one or two examples for this sentence. Imagine someone using your voice for like, hate speech or something and that permanently affecting your career or social life.
The context it exists in. Ultimately they're both just neutral tools, but AI art generators and the culture around it had malicious intent towards artists (to the point that there are people purposely targeting "No AI" artists). Diff-svc on the other hand, it's still too early to tell, but currently the culture around diff-svc is just people toying around with it and putting their own data into it. In the context of vocal synth fandom, its just another medium to put their own characters in like NNSVS or UTAU. If however people start using diff-svc for malicious reasons, the you can expect people to quickly turn against it like back when people turned against AI art generators when it stopped being a fun thing to toy around with and started becoming malicious towards artists.
While yes, they are the same act, there is a vital factor that changes everything about its ethics: consent. AI art generators are automatically trained on all publically available art, regardless of whether the artist wants to participate. All the user has to do is type in the name of an artist, and the technology will replicate their work. This violates just about every principle of ethical labour: the person whose work is being used must give permission and be compensated for their contribution. The vocal version needs that data to be added manually and by choice, which most people are doing with their own voices. Consent changes it from being fun and interesting to violating.
Now this does mean that any vocal synth created from someone's voice without their consent is unethical, but that includes any technology, from the oldest UTAUs to silly 'I made an AI sing this in the style of Eminem' memes. So its not an issue inherant or specific to AI.
ive thought more and more about vocsynth AI lately too with all the AI art stuff and i wondered when we might get to a place where tools were getting similar to AI art generation. theres two AI generation things i dislike and make me uncomfy with their uses: AI art, for the much explained reasons of art theft/lack of consent/etc AI text to speech you find online (the ones full of characters that have been recreated) which i dislike for similar no consent reasons and the availability of the tools to the public (training/doing it all yourself for a meme vs the tool being readily available with an abundance of options for anyone to use for whatever purpose meme or not- the difference between personal/private use and public).
i agree its the context that can shift the moral rightness-- for one i think its good (and i would hope it stays this way) that diff-svc is a tool where you train your own data to get your result instead of an already packaged library of data and voices built off of data that may or may not have permission for which is totally what loads of art AI does. and i have seen people already make memey banks for characters-- this would again bring up the morality of it mostly when you look to those AI TTS (which i dislike). and i think its clear this can never be a black and white thing-- because when i think about even using diff-svc to recreate voices that arent you own, my general thought is yeah, this can be used for bad obviously. though any AI can (and id while vocalochange isnt making a new AI bank, its a similar issue too of ripping and covering vocals that arent your own too) and i think what i dont like is ease of access/availability in the form of "we did all the training and stuff for you, using art that we dont own, now you can make more art off of it that isnt your own!".
im even saying this as someone who teeters on that moral choice of recreating voices that arent yours, i made utau banks out of sonic the hedgehog and totally understand the backs and forths on doing that-- but my banks arent available, theyre not just out there for anyone to just starting covering/making them say whatever they want. for the reasons of utau's own terms but also the fact that like. you want to make someone say something stupid? i guess id rather you go through the hard work yourself if you want to do it that bad, than me just metaphorically handing you the keys to my car so you can crash it two blocks down the street. im not sure if that all adds up but i think about this quite a bit due to those utau banks i made and the fact i have strong uncomfortable feelings about AI audio/voice done without consent-- the idea of someone recreating your voice without consent in a way that can be so realistic its hard to tell?? no thanks. even further, providing it to the public? hell no.
conclusion: context does matter in any case and i think for me personally the raw "you need to do all the work yourself" is an important boundary, when AI art generation simply has been trained and allows you to recreate whatever you please without needing to do that training work. the training itself can be ethical if you are the one training it, and you train it off your own data, and the ease of access can cause more moral problems because of the lack of permission or consent. even further the casual availability and access (go to a web and just clicking on some AI tts) leaves little questioning from the consumers point, they have no context on the technological side, and it will always remain a mystery to them even more so when the sites/groups hosting these tools fail to ever give any more information about it (sadly leaving many totally unaware on how this stuff works, or to even question it). when users alone decide to train something its their own choice of actions (to do something good or bad) and not the company/faceless entity providing morally grey tools prepared for everyone else (where the users' lack of knowledge may have them committing to an act they may not fully understand). hope that didnt go too off the rails lol, AI generation morality gets a bit general when you start to go into it.
i agree its the context that can shift the moral rightness-- for one i think its good (and i would hope it stays this way) that diff-svc is a tool where you train your own data to get your result instead of an already packaged library of data and voices built off of data that may or may not have permission for which is totally what loads of art AI does. and i have seen people already make memey banks for characters-- this would again bring up the morality of it mostly when you look to those AI TTS (which i dislike). and i think its clear this can never be a black and white thing-- because when i think about even using diff-svc to recreate voices that arent you own, my general thought is yeah, this can be used for bad obviously. though any AI can (and id while vocalochange isnt making a new AI bank, its a similar issue too of ripping and covering vocals that arent your own too) and i think what i dont like is ease of access/availability in the form of "we did all the training and stuff for you, using art that we dont own, now you can make more art off of it that isnt your own!".
im even saying this as someone who teeters on that moral choice of recreating voices that arent yours, i made utau banks out of sonic the hedgehog and totally understand the backs and forths on doing that-- but my banks arent available, theyre not just out there for anyone to just starting covering/making them say whatever they want. for the reasons of utau's own terms but also the fact that like. you want to make someone say something stupid? i guess id rather you go through the hard work yourself if you want to do it that bad, than me just metaphorically handing you the keys to my car so you can crash it two blocks down the street. im not sure if that all adds up but i think about this quite a bit due to those utau banks i made and the fact i have strong uncomfortable feelings about AI audio/voice done without consent-- the idea of someone recreating your voice without consent in a way that can be so realistic its hard to tell?? no thanks. even further, providing it to the public? hell no.
conclusion: context does matter in any case and i think for me personally the raw "you need to do all the work yourself" is an important boundary, when AI art generation simply has been trained and allows you to recreate whatever you please without needing to do that training work. the training itself can be ethical if you are the one training it, and you train it off your own data, and the ease of access can cause more moral problems because of the lack of permission or consent. even further the casual availability and access (go to a web and just clicking on some AI tts) leaves little questioning from the consumers point, they have no context on the technological side, and it will always remain a mystery to them even more so when the sites/groups hosting these tools fail to ever give any more information about it (sadly leaving many totally unaware on how this stuff works, or to even question it). when users alone decide to train something its their own choice of actions (to do something good or bad) and not the company/faceless entity providing morally grey tools prepared for everyone else (where the users' lack of knowledge may have them committing to an act they may not fully understand). hope that didnt go too off the rails lol, AI generation morality gets a bit general when you start to go into it.
Last edited:
You don't even really have to wait to see this is the thing, you just have to step outside the vocal synth fandom for a bit. I'm pretty sure I've talked before in this very thread about the incredibly negative reaction social media had to Disney using AI to recreate a young Mark Hamill's voice in (I believe?) The Mandalorian. And more recently, I read news about some video game developers using (or just considering using) virtual voice actors for their NPCs, and the reaction I saw to that was even more incensed. Like, everything from a (to be fair, not entirely unfounded) fear of real human voice actors becoming obsolete in the near future, to an almost spiritual revulsion to the very concept of synthesized voice acting; I'm talking Hayao Miyazaki "I strongly feel this is an insult to life itself" kind of stuff.
What I'm trying to get at is, a lot of people already see AI-based voice synthesis in exactly the same way that many people now feel about "AI-generated art", including many of us vocal synth fans. And part of me wonders if we are in the exact same kind of bubble that the pro-AI art folks are in, and if there isn't a bit of hypocrisy in us vehemently denouncing the likes of NovelAI for sampling the works of unwilling artists on Pixiv or DeviantART... and then turning around to make Diff-SVC banks for Peppa Pig or Beyonce or whoever. If the absence or presence of consent on the part of the person/artist being "cloned" (for lack of a better term) is the main thing that makes it ethical or unethical (and I'm inclined to agree with that, personally!), then all those "silly" celebrity/character voicebanks are every bit as unethical as those image generating machines.
Anyway I guess my real Unpopular Opinion in all this is that when it really comes down to it, it's not a technological issue at all, despite how it's usually framed in discussions; It's a labor issue. If your ultimate fear is about artists, voice actors, and other creatives becoming obsolete at the hands of new technology, your fear isn't really about the technology itself but about the society that it exists in. It's the fact that there are ultra-wealthy powerful people that can and will do anything to get as much profit for the least effort, and who will gladly condemn countless humans to poverty and misery if it'll help them achieve that goal. Machine learning is just one of many tools available to them; if it didn't exist, they'd just look elsewhere. But machine learning and other AI techniques also have plenty of legitimate, or at the very least harmless, uses that I think people getting mad at the tech itself are missing the forest for the trees extremely hard.
I guess, tl;dr... I wish folks would be a little less mad at people using AI to make shitposts (or even maybe, occasionally, legitimately interesting art?) and a little more mad that we inhabit an economic system where countless people are at risk of losing their livelihoods in an instant with little hope of recourse. For all the digital ink that's been spilled about it recently, stable diffusion is pretty low down on the list of Existential Risks to Society. In my opinion, anyway.
I probably sound like a biased Cevio liker at this point but after the amount of takes I’ve read on the engine & its voices (these mostly being on the cursed blue bird app so I shouldn’t even read that deeply into it) I keep wondering if half the ppl who are extremely critical about it even listened to more than half a demo of its voices??? Claims about the Kamitsubaki voices being appends of each other, every other vocal being Sasara appends.. very specifically I never understood the Kafu and Sekai comparisons, or actually I do get it a bit cuz the higher pitch from the base alpha parameter already sounding formant shifted mixed with the current engine noise can make it seem a bit like that in Cevio BUT they still sing in completely different styles… they honestly emulate their respective VPs’ singing style pretty well (and Kaf & Isekaijoucho have pretty distinct voices from each other as well). This aside the Alpha parameter also exists and it’s rly only nerfed for Kafu and increasing it for Sekai makes her sound a hell of a lot closer to Ojou. I also find it a bit funny that a lot of this also comes from people who eat up every single SynthV letter of the week who also have a very apparent same vocal type syndrome going on despite the engine noise being different but anyways I don’t wanna start on this fkdkslfkg
Going back to the singing style point, I think it is the biggest plus Cevio AI has over other current AI singing synths: it does a pretty damn good job at retaining singing styles and quirks from the voices their banks are based on. IA AI may have been clowned a bit on this end of the fandom and her initial showings didn’t do her justice but I still find it impressive how close she can sound to Lia just out of the box… she may be one of the trickiest banks to work with cause her AI tuning can be pretty wild but I also commend how it commits to keeping it close to Lia’s singing style. The other Cevio AI banks I’ve tried aren’t as much of a challenge to work with but I find it really fun to work with these singing styles and tuning them to fit into different songs! The forced auto-tuning may not be for everyone and I understand that, but ever since I’ve read Cevio banks being described as “giving a singer a script while you act as their vocal director” as opposed to being more of instruments you’re in control of like Vocaloid/SynthV with auto tuning off/Utau I kind of get the vision for Cevio AI a lot better now lol. I can see why this can turn ppl off who prefer a more neutral base that other synths can give you to work with, but I honestly find it makes using Cevio AI really fun to use to me like I said earlier and the AI part of it doesn’t really eliminate the user input when it comes to tuning.
Another really weird take I’ve seen is people claiming SynthV not being popular in Japan is due to… xenophobia??? Which I honestly find baffling, please correct me if I’m wrong but these seem like baseless assumptions and I’ll retract this if there’s anything pointing to that. It’s true something like this may have happened with SeeU back in 2011 but I really don’t see any indication that anything similar is happening here. This reads a bit like denial coming from some people about SynthV just not being as popular as engines that they deem as inferior lol. Fact is that marketing is minimal and not that great outside of their Chinese market & there just isn’t much to write home about when it comes to their Japanese output… maybe up until Mai’s release which seems to have attracted quite a bit of attention for once! Which is proof to me that their lack of success so far isn’t derived from prejudice but from there not really being much of an attractive offering or incentives to buy the program up till now.
As impressive as SV can be the sales aren’t gonna form out of thin air if it has no worthwhile exposure
Mai is a standout among the alphabet soup and a free voice for SynthV Pro buyers who had a pretty impressive sounding demo that got quite a decent bit of views and the result of that is SV Pro being on top of DLSite sale rankings rn. See how that works!!!!! It’s finally a step in the right direction and hopefully one that they’ll keep up. It doesn’t help that the whole 1.7 fiasco kind of fumbled the bag a bit when it comes to 3rd party and I wonder how sustainable the feature creep every single update can be to get older VBs up to date but hopefully something like the HDVM situation doesn’t happen again…
anyways sorry I went on a bit of a rant here but I also had some of this stuff on my mind for a bit and I wanted to get some of it out
Going back to the singing style point, I think it is the biggest plus Cevio AI has over other current AI singing synths: it does a pretty damn good job at retaining singing styles and quirks from the voices their banks are based on. IA AI may have been clowned a bit on this end of the fandom and her initial showings didn’t do her justice but I still find it impressive how close she can sound to Lia just out of the box… she may be one of the trickiest banks to work with cause her AI tuning can be pretty wild but I also commend how it commits to keeping it close to Lia’s singing style. The other Cevio AI banks I’ve tried aren’t as much of a challenge to work with but I find it really fun to work with these singing styles and tuning them to fit into different songs! The forced auto-tuning may not be for everyone and I understand that, but ever since I’ve read Cevio banks being described as “giving a singer a script while you act as their vocal director” as opposed to being more of instruments you’re in control of like Vocaloid/SynthV with auto tuning off/Utau I kind of get the vision for Cevio AI a lot better now lol. I can see why this can turn ppl off who prefer a more neutral base that other synths can give you to work with, but I honestly find it makes using Cevio AI really fun to use to me like I said earlier and the AI part of it doesn’t really eliminate the user input when it comes to tuning.
Another really weird take I’ve seen is people claiming SynthV not being popular in Japan is due to… xenophobia??? Which I honestly find baffling, please correct me if I’m wrong but these seem like baseless assumptions and I’ll retract this if there’s anything pointing to that. It’s true something like this may have happened with SeeU back in 2011 but I really don’t see any indication that anything similar is happening here. This reads a bit like denial coming from some people about SynthV just not being as popular as engines that they deem as inferior lol. Fact is that marketing is minimal and not that great outside of their Chinese market & there just isn’t much to write home about when it comes to their Japanese output… maybe up until Mai’s release which seems to have attracted quite a bit of attention for once! Which is proof to me that their lack of success so far isn’t derived from prejudice but from there not really being much of an attractive offering or incentives to buy the program up till now.
As impressive as SV can be the sales aren’t gonna form out of thin air if it has no worthwhile exposure
Mai is a standout among the alphabet soup and a free voice for SynthV Pro buyers who had a pretty impressive sounding demo that got quite a decent bit of views and the result of that is SV Pro being on top of DLSite sale rankings rn. See how that works!!!!! It’s finally a step in the right direction and hopefully one that they’ll keep up. It doesn’t help that the whole 1.7 fiasco kind of fumbled the bag a bit when it comes to 3rd party and I wonder how sustainable the feature creep every single update can be to get older VBs up to date but hopefully something like the HDVM situation doesn’t happen again…anyways sorry I went on a bit of a rant here but I also had some of this stuff on my mind for a bit and I wanted to get some of it out
i agree on like every point that u've made dwdffdsfxdsfxdsfdxs the people on the blue bird app do criticize cevio way too much when compared to synthv, and tbh there are a lot of reoccurring statements they make that do kind of hint at the fact that they haven't really listened to cevio that much (aka the way people are saying that there are kafu clones on cevio which is like? the only bank that sounds like kafu is rime, and she was defenitly supposed to be like kafu but more mature, which in my opinion she does really well with)I probably sound like a biased Cevio liker at this point but after the amount of takes I’ve read on the engine & its voices (these mostly being on the cursed blue bird app so I shouldn’t even read that deeply into it) I keep wondering if half the ppl who are extremely critical about it even listened to more than half a demo of its voices??? Claims about the Kamitsubaki voices being appends of each other, every other vocal being Sasara appends.. very specifically I never understood the Kafu and Sekai comparisons, or actually I do get it a bit cuz the higher pitch from the base alpha parameter already sounding formant shifted mixed with the current engine noise can make it seem a bit like that in Cevio BUT they still sing in completely different styles… they honestly emulate their respective VPs’ singing style pretty well (and Kaf & Isekaijoucho have pretty distinct voices from each other as well). This aside the Alpha parameter also exists and it’s rly only nerfed for Kafu and increasing it for Sekai makes her sound a hell of a lot closer to Ojou. I also find it a bit funny that a lot of this also comes from people who eat up every single SynthV letter of the week who also have a very apparent same vocal type syndrome going on despite the engine noise being different but anyways I don’t wanna start on this fkdkslfkg
Going back to the singing style point, I think it is the biggest plus Cevio AI has over other current AI singing synths: it does a pretty damn good job at retaining singing styles and quirks from the voices their banks are based on. IA AI may have been clowned a bit on this end of the fandom and her initial showings didn’t do her justice but I still find it impressive how close she can sound to Lia just out of the box… she may be one of the trickiest banks to work with cause her AI tuning can be pretty wild but I also commend how it commits to keeping it close to Lia’s singing style. The other Cevio AI banks I’ve tried aren’t as much of a challenge to work with but I find it really fun to work with these singing styles and tuning them to fit into different songs! The forced auto-tuning may not be for everyone and I understand that, but ever since I’ve read Cevio banks being described as “giving a singer a script while you act as their vocal director” as opposed to being more of instruments you’re in control of like Vocaloid/SynthV with auto tuning off/Utau I kind of get the vision for Cevio AI a lot better now lol. I can see why this can turn ppl off who prefer a more neutral base that other synths can give you to work with, but I honestly find it makes using Cevio AI really fun to use to me like I said earlier and the AI part of it doesn’t really eliminate the user input when it comes to tuning.
Another really weird take I’ve seen is people claiming SynthV not being popular in Japan is due to… xenophobia??? Which I honestly find baffling, please correct me if I’m wrong but these seem like baseless assumptions and I’ll retract this if there’s anything pointing to that. It’s true something like this may have happened with SeeU back in 2011 but I really don’t see any indication that anything similar is happening here. This reads a bit like denial coming from some people about SynthV just not being as popular as engines that they deem as inferior lol. Fact is that marketing is minimal and not that great outside of their Chinese market & there just isn’t much to write home about when it comes to their Japanese output… maybe up until Mai’s release which seems to have attracted quite a bit of attention for once! Which is proof to me that their lack of success so far isn’t derived from prejudice but from there not really being much of an attractive offering or incentives to buy the program up till now.
As impressive as SV can be the sales aren’t gonna form out of thin air if it has no worthwhile exposure
anyways sorry I went on a bit of a rant here but I also had some of this stuff on my mind for a bit and I wanted to get some of it out
based blackoutI probably sound like a biased Cevio liker at this point but after the amount of takes I’ve read on the engine & its voices (these mostly being on the cursed blue bird app so I shouldn’t even read that deeply into it) I keep wondering if half the ppl who are extremely critical about it even listened to more than half a demo of its voices??? Claims about the Kamitsubaki voices being appends of each other, every other vocal being Sasara appends.. very specifically I never understood the Kafu and Sekai comparisons, or actually I do get it a bit cuz the higher pitch from the base alpha parameter already sounding formant shifted mixed with the current engine noise can make it seem a bit like that in Cevio BUT they still sing in completely different styles… they honestly emulate their respective VPs’ singing style pretty well (and Kaf & Isekaijoucho have pretty distinct voices from each other as well). This aside the Alpha parameter also exists and it’s rly only nerfed for Kafu and increasing it for Sekai makes her sound a hell of a lot closer to Ojou. I also find it a bit funny that a lot of this also comes from people who eat up every single SynthV letter of the week who also have a very apparent same vocal type syndrome going on despite the engine noise being different but anyways I don’t wanna start on this fkdkslfkg
Going back to the singing style point, I think it is the biggest plus Cevio AI has over other current AI singing synths: it does a pretty damn good job at retaining singing styles and quirks from the voices their banks are based on. IA AI may have been clowned a bit on this end of the fandom and her initial showings didn’t do her justice but I still find it impressive how close she can sound to Lia just out of the box… she may be one of the trickiest banks to work with cause her AI tuning can be pretty wild but I also commend how it commits to keeping it close to Lia’s singing style. The other Cevio AI banks I’ve tried aren’t as much of a challenge to work with but I find it really fun to work with these singing styles and tuning them to fit into different songs! The forced auto-tuning may not be for everyone and I understand that, but ever since I’ve read Cevio banks being described as “giving a singer a script while you act as their vocal director” as opposed to being more of instruments you’re in control of like Vocaloid/SynthV with auto tuning off/Utau I kind of get the vision for Cevio AI a lot better now lol. I can see why this can turn ppl off who prefer a more neutral base that other synths can give you to work with, but I honestly find it makes using Cevio AI really fun to use to me like I said earlier and the AI part of it doesn’t really eliminate the user input when it comes to tuning.
Another really weird take I’ve seen is people claiming SynthV not being popular in Japan is due to… xenophobia??? Which I honestly find baffling, please correct me if I’m wrong but these seem like baseless assumptions and I’ll retract this if there’s anything pointing to that. It’s true something like this may have happened with SeeU back in 2011 but I really don’t see any indication that anything similar is happening here. This reads a bit like denial coming from some people about SynthV just not being as popular as engines that they deem as inferior lol. Fact is that marketing is minimal and not that great outside of their Chinese market & there just isn’t much to write home about when it comes to their Japanese output… maybe up until Mai’s release which seems to have attracted quite a bit of attention for once! Which is proof to me that their lack of success so far isn’t derived from prejudice but from there not really being much of an attractive offering or incentives to buy the program up till now.
As impressive as SV can be the sales aren’t gonna form out of thin air if it has no worthwhile exposure
anyways sorry I went on a bit of a rant here but I also had some of this stuff on my mind for a bit and I wanted to get some of it out
I'm gonna steal this from @pico as I've seen them mention it on twitter, forgive me
but I also don't like the attitude of "why would they develop for these objectively inferior engines 
" completely overlooking how truly amazing it is how Techno Speech and Kamitsubaki were able to communicate and go "Yup 
we decided to implement the emotion slider to truly realize Sekai's potential! Btw from here on out emotion slider for all of you 
" and from what I understand they're developing greater support for Haru's rapping? I don't fully get it so please do educate me, my feelings on Kamitsubaki are complicated but how the two studios were able to communicate and make the engine even better in collaborative effort or how they're willing to customize things for vbs can not go underappreciated!On the other end, you had the whole HDVM fiasco, where it was pretty much revealed that 3rd party developers are left in the dark, all 3rd party developers were first to apologize and act, leaving DT to be one of the last ones to react, the development on the matter's been very messy and sluggish too, and I don't believe the beta releases, which are basically are way to prevent disasters like this to happen in the future, would've even happened if DT wouldn't have gotten slap on the wrist from Tokyo6 ( which is honestly embarrassing ), and even then, miscommunications still appear to be an issue considering what's happened with Chifuyu's Lite vb and Eleanor's beta vb, and the feature creep is still an issue, leaving the future of vbs on the engine uncertain.
It's just all really really bad look when you contrast it next to techno speech, and when put side by side it's no wonder Cevio AI is more favored in terms of development.
I've seen excuses for it made already that it's just their "1st little oopsie
" but that's in my opinion grossly underestimating just how bad all mentioned above really is...and that HDVM on launch was a result of an issue that's been going on this whole time and only became most apparent with HDVM.
The GENBU AI situation also comes to mind... I kind of remember the general vibe around that at fist being that "ah, the small doujin circle was overexcited and made a huge clerical error" but after reading their full explanation after the dust had cleared, it felt to me more like even though it was VirVox's mistake in announcing the voicebank that caused the project to dissolve, the 3rd party ecosystem for the engine seemed unexpectedly 'grassroots', a feeling that I think was strengthened after revelations involving HDVM, 3rd party vocal modes launching with varying degrees of success, and SynthesizerV's other model implementations.
I think there is some truth to the idea in some recent birdsite posts about CeVIO that CeVIO and Techno-Speech's singular focus on the Japanese market is a drawback. Despite having an English UI, there is no official English translation available for CeVIO's online resources. I'm also not sure about the product's documentation, my assumption is that it's also in Japanese, but that might be wrong, so don't quote me on that.
It certainly seems prescient that CeVIO has been incredibly effective at focusing on the market of "doujin" music creators in a way Yamaha did not and Dreamtonics certainly never will. I think this is a big part of why English-speaking fans often don't have as big of an awareness of CeVIO and its progress over the years until CeVIO AI suddenly and massively exploded in the public consciousness after the pretty much entirely unexpected success of KAFU. The CeVIO characters were not really uncommon characters at all before KAFU, but they were barely discussed by english speakers (IMO).
Techno-Speech seems to realize that CeVIO as a product will be unable to be separated from the niche otaku culture that fostered it, which is why I personally believe CeVIO Pro was quickly rebranded as VoiSona and is quite different in its presentation.
In contrast, SynthesizerV's early focus on English-speaking markets and the success of AI synthesis itself for English speakers immediately made it popular among our small (probably growing?) demographic. So with this stuff in mind I completely understand and empathize with why English speakers are so passionate about SynthesizerV and I'm really glad there are groups that have spawned from the SynthesizerV community like VirVox and Eclipsed Sounds that truly have the interests of these fans at heart. But it really does go to show that nothing is perfect, not even SynthesizerV, and there is real value in experimenting with lots of software, different kinds of technology, and embracing the different facets of our community that create good and do good work. (Plus, all of the doujin-y stuff has real heart, and is often not dissimilar to the culture of "old vocaloid" that so many fans regularly lament is gone/they missed out of, even though the spirit is very much alive and exists in different forms at this very moment, even if it isn't the sudden spike of explosive popularity that existed in the 2000's and early 2010's.)
The introduction of Mai hasn't been the only spike in awareness for SynthesizerV in the Japanese-speaking sphere either, if my memory serves. I remember when cillia published a cover of a song (it may have been the Higurashi cover? would need to really go back and check) with Saki AI, it was at the top of the niconico music rankings for ages. Japanese users are very much willing to embrace SynthesizerV but it's far from the only good option in their corner, and as has been discussed in the SynthV thread is not the engine's only drawback. This is more than enough to explain why SynthesizerV is not the monolith among Japanese fans that it is among english-speaking fans. And that's OK!!
There's no use in creating a 'console wars' culture around our software when the userbases and even the products themselves are widely shared between these corporate entities and it's been proven over and over again that problems will arise no matter what camp you pick. There was still drama and licensing issues and disappointment when VOCALOID was the only commercial singing synth anyone was excited about. And especially in today's sphere, we have so many good options available to us. CeVIO is an engine with unique advantages as is SynthesizerV is an engine with unique advantages. As a community comprised with artists or otherwise explicitly concerned with art, it would do us all good to give lots of possibilities the time of day-- comparatively very few people worldwide were sold on KAFU when the final voice was announced, but look where it is now...? (And yet I've seen widely shared sentiments that KAFU as an example [it's a really good example for a lot of things...] does not have extremely wide popularity and was not a seriously commercially successful product. Huh?!)
I think the community has an incredibly bright future ahead and we have so much to be excited about, but it would be great if there was wider cognition of the community and the products as a whole.
I think there is some truth to the idea in some recent birdsite posts about CeVIO that CeVIO and Techno-Speech's singular focus on the Japanese market is a drawback. Despite having an English UI, there is no official English translation available for CeVIO's online resources. I'm also not sure about the product's documentation, my assumption is that it's also in Japanese, but that might be wrong, so don't quote me on that.
It certainly seems prescient that CeVIO has been incredibly effective at focusing on the market of "doujin" music creators in a way Yamaha did not and Dreamtonics certainly never will. I think this is a big part of why English-speaking fans often don't have as big of an awareness of CeVIO and its progress over the years until CeVIO AI suddenly and massively exploded in the public consciousness after the pretty much entirely unexpected success of KAFU. The CeVIO characters were not really uncommon characters at all before KAFU, but they were barely discussed by english speakers (IMO).
Techno-Speech seems to realize that CeVIO as a product will be unable to be separated from the niche otaku culture that fostered it, which is why I personally believe CeVIO Pro was quickly rebranded as VoiSona and is quite different in its presentation.
In contrast, SynthesizerV's early focus on English-speaking markets and the success of AI synthesis itself for English speakers immediately made it popular among our small (probably growing?) demographic. So with this stuff in mind I completely understand and empathize with why English speakers are so passionate about SynthesizerV and I'm really glad there are groups that have spawned from the SynthesizerV community like VirVox and Eclipsed Sounds that truly have the interests of these fans at heart. But it really does go to show that nothing is perfect, not even SynthesizerV, and there is real value in experimenting with lots of software, different kinds of technology, and embracing the different facets of our community that create good and do good work. (Plus, all of the doujin-y stuff has real heart, and is often not dissimilar to the culture of "old vocaloid" that so many fans regularly lament is gone/they missed out of, even though the spirit is very much alive and exists in different forms at this very moment, even if it isn't the sudden spike of explosive popularity that existed in the 2000's and early 2010's.)
The introduction of Mai hasn't been the only spike in awareness for SynthesizerV in the Japanese-speaking sphere either, if my memory serves. I remember when cillia published a cover of a song (it may have been the Higurashi cover? would need to really go back and check) with Saki AI, it was at the top of the niconico music rankings for ages. Japanese users are very much willing to embrace SynthesizerV but it's far from the only good option in their corner, and as has been discussed in the SynthV thread is not the engine's only drawback. This is more than enough to explain why SynthesizerV is not the monolith among Japanese fans that it is among english-speaking fans. And that's OK!!
There's no use in creating a 'console wars' culture around our software when the userbases and even the products themselves are widely shared between these corporate entities and it's been proven over and over again that problems will arise no matter what camp you pick. There was still drama and licensing issues and disappointment when VOCALOID was the only commercial singing synth anyone was excited about. And especially in today's sphere, we have so many good options available to us. CeVIO is an engine with unique advantages as is SynthesizerV is an engine with unique advantages. As a community comprised with artists or otherwise explicitly concerned with art, it would do us all good to give lots of possibilities the time of day-- comparatively very few people worldwide were sold on KAFU when the final voice was announced, but look where it is now...? (And yet I've seen widely shared sentiments that KAFU as an example [it's a really good example for a lot of things...] does not have extremely wide popularity and was not a seriously commercially successful product. Huh?!)
I think the community has an incredibly bright future ahead and we have so much to be excited about, but it would be great if there was wider cognition of the community and the products as a whole.
All good! :)I'm gonna steal this from @pico as I've seen them mention it on twitter, forgive me
You put it perfectly with not wanting to promote a "console wars". SynthV Twitter has a problem with this, since they shit on CeVIO and Vocaloid for not being SynthV, and try to force SynthV on Emvoice users (I know Emvoice is a touchy subject, especially for us weebs, but I still find it unfair that when someone asks for help setting it up on Twitter, they are bombarded with "just use SynthV instead"). But it's a great thing that so many options exist now because it forces these companies to share the vocal market. They all come with different ideas, workflows, and audiences.
I also find it a bit funny that a lot of this also comes from people who eat up every single SynthV letter of the week who also have a very apparent same vocal type syndrome going on despite the engine noise being different but anyways I don’t wanna start on this fkdkslfkg
...
It doesn’t help that the whole 1.7 fiasco kind of fumbled the bag a bit when it comes to 3rd party and I wonder how sustainable the feature creep every single update can be to get older VBs up to date but hopefully something like the HDVM situation doesn’t happen again…
anyways sorry I went on a bit of a rant here but I also had some of this stuff on my mind for a bit and I wanted to get some of it out
This may be completely anectodal to the specific people I follow and see and social media - but I find this opinion isn't unpopular at all. People who love SynthV are generally sick of the alphabet soup and unhappy with how Dreamtonics deals with third party voicebanks. The vast majority of SynthV Twitter was clowning on Dreamtonics for their HDVM update. We like how SynthV sounds, which is why Dreamtonics' choices hurt so much.On the other end, you had the whole HDVM fiasco, where it was pretty much revealed that 3rd party developers are left in the dark, all 3rd party developers were first to apologize and act, leaving DT to be one of the last ones to react, the development on the matter's been very messy and sluggish too, and I don't believe the beta releases, which are basically are way to prevent disasters like this to happen in the future, would've even happened if DT wouldn't have gotten slap on the wrist from Tokyo6 ( which is honestly embarrassing ), and even then, miscommunications still appear to be an issue considering what's happened with Chifuyu's Lite vb and Eleanor's beta vb, and the feature creep is still an issue, leaving the future of vbs on the engine uncertain.
It's just all really really bad look when you contrast it next to techno speech, and when put side by side it's no wonder Cevio AI is more favored in terms of development.
I've seen excuses for it made already that it's just their "1st little oopsie
To Dreamtonics' credit, they seemed to have finally hired people to do their marketing, and SynthV is starting to reach out their intended audience.
But just as it was mentioned, I wish Dreamtonics allowed AHSoftware and other third party companies to branch out into the Japanese doujin community more. It is possible, with multiple SynthV voicebanks that have already been appreciated by Japanese users. I completely agree that some SynthV users are way too harsh on CeVIO, when it's catering to the same fanbase that love(d) Vocaloid.
I imagine high-pitched, formant-shifted voices like Kafu and Sekai are not going to appeal to a lot of those who like the very natural sound of SynthV (but of course there are multiple people who like both kinds of voices). Kafu also has an issue of being really muffled. But voices like Coko, Poppy and Rose show that CeVIO is equally capable of capturing voices that feel real.
I think most people who think SynthV is objectively superior just don't like the artificial sound of CeVIO's engine. It's not a problem when the voice is strong and belting, but it really jumps out in softer voices. Some people actually like the engine noise, though, and that's cool. Hearing how good Coko and Rose's belting was made me really excited for c flower.
But inversely to SynthesizerV, CeVIO doesn't cater to countries outside of Japan at all. I would be more excited for Rose and c flower if they had proper cross-lingual support. Hopefully, one day! I also realise this post doesn't really contain any unpopular opinions either, so I'm not telling anyone off lol. It's just nice to talk about this, because the bird app does not allow much nuance in discussions.
Personally, I think the argument is less “People act like SynthV can do no wrong,” which, as you’ve pointed out, isn’t quite the case these days, and more “Even at its most criticized, people are still willing to overlook SynthV’s flaws in favor of its strengths, something CeVIO isn’t afforded in this side of the fandom.” Each engine has strengths and weaknesses and different appeals, but there’s still such a huge attitude of negativity towards anything that’s not SynthV. Someone made a comment on VocaTwitter in response to SeeU and Uni’s possible returns saying they personally hoped they were on CeVIO rather than SynthV, and someone replied “When you hate quality and love engine noise.” It had about three times the amount of likes as the original comment.
Maybe I’m just jaded because of how toxic the SynthV community has been in regards to any synth that’s not it, and I acknowledge that bias. I just wish CeVIO was afforded even a fraction of the good faith SynthV is given over here. Same can be said for any synth that gets clowned on for not being SynthV, but CeVIO gets the worst of it without a doubt.
I didn't share these critiques in isolation, but as contrast to Cevio's strengths, strengths that get overlooked, it's to highlight the refusal of the community to expand onto another editor ( highlight on the word expand, not move to ) despite it doing things that are desired for in the community ( focus on characters and preserving the doujin-y feel ), or not even expand onto but to be even given recognition for it.
Which is just expanding on the sentiment shared in the thread, that instead of picking camps, it's the most productive to embrace one engine's strengths where the other lacks, in current's climate however, even at the peak of SVS receiving more and more criticism, the reaction to offering Cevio as another option is met with "but it's poo-poo pee-pee"
Also what Peachy said
And with anecdotal evidence on my end, even when criticized, I find that the community is a lot softer with it's criticism towards SVS, I've seen the community be at it's most polite when directing their criticism at it, which says a lot in my book considering they're capable of borderline harassment for a lot less, I've seen people get more worked up over V5 going on sale than the status of Eleanor's boxes.
Apologies for double posting but since I mentioned Vocaloid I got reminded of another 🧅 of mine:
I'm actually glad V6's release was a surprise, for my own sanity, because I'm certain that if we would've gotten announcement pertaining to it prior to it's release, people would have made up their opinions on it prematurely and gotten themselves worked up and then actually would've approached V6 absolutely close mindedly on launch, people were showing tendencies to do so but releasing it this way put a stop to the worst of it I think.
Utterly speculative on my part but I think a lot of the complaints coming from people who typically approach Vocaloid in bad faith criticize the lack of prior announcements less so as genuine criticism of the marketing, but because they didn't get a room to "make up a guy and then get mad at him".
I am miffed about people creating Diff-SVC voicebanks from people who haven't consented to it. Consent is already such a huge issue in AI technology. A couple of covers made with Peppa Pig's voice may seem harmless, but I've seen voicebanks made from CupcakKe and Billie Eilish's voices as well. At least in UTAU, you could clearly hear the stolen voice was artificial and rough due to a lack of appropriate samples. Now, even though the creators don't share these voicebanks, it still violates consent while creating pretty convincing models out of real people's voices. It doesn't feel right. I'm worried someone will get a hang of what's going on and ruin this really cool way of making a voice database.
On a completely different topic, I saw some folks comment on Twitter that Asterian's voice is too niche. Nothing could be the furthest thing from the truth, but a lot of people just don't know how to bring out colour and expression in baritone and bass voices. They're so rare. I hope Asterian will grow in popularity despite this.
On a completely different topic, I saw some folks comment on Twitter that Asterian's voice is too niche. Nothing could be the furthest thing from the truth, but a lot of people just don't know how to bring out colour and expression in baritone and bass voices. They're so rare. I hope Asterian will grow in popularity despite this.
Wait but didn't the SynthV community WANT niche voicebanks? Even then, from his vocal mode demo he seems almost as versatile as Solaria.