The OLIVER demo sounded good to me. I agree, I am just baffled as to why all of this was dripped out over time instead of delivered to potential backers as a more cohesive look at what was to come if the project came to fruition. Compared to what has surfaced in the last couple of weeks, the original campaign is hilariously underwhelming. Continuing to be confused as to why the IndieGoGo was handled the way it was, but pleased with recent demo content.
This might be old-news but the voicebanks that people thought were Sweet Ann and Big Al have confirmed names.
Glassia and Gabriel.
The completionist bundle is also currently on sale at 33% off!
Glassia and Gabriel.
The completionist bundle is also currently on sale at 33% off!
The Sweet Ann and Big Al banks are actually these ones! The top one is probably Ruby.
A bit worried that Vocatone hasn't posted an update about the Aurum demo that was supposed to be last month.
A bit worried that Vocatone hasn't posted an update about the Aurum demo that was supposed to be last month.
OOPs sorry. I remembered the red color but not which silohettes they were. Thank you for correcting me!!! 
Hopefully it comes out soon? Considering they opted to show the VAs demos instead of the banks- i'm assuming it's still very in-progress.
Hopefully it comes out soon? Considering they opted to show the VAs demos instead of the banks- i'm assuming it's still very in-progress.
Oh god please be sweet Ann the design is reminiscent of the popular woc design that I am a fan of! Here’s hoping
i can tell the "sweet ann" vocalist design has textured hair :D so excited if it ends up not being so i owe everyone 5 dollars
You can tell us if Big Al's voice provider is the same as his Vocaloid? Because they had to record everything again with a new voice provider and, honestly, I really love Big Al's voice XD I hope they called the same voice provider (Tho would love to have another VB with the first intended VA).The Sweet Ann and Big Al banks are actually these ones! The top one is probably Ruby.
A bit worried that Vocatone hasn't posted an update about the Aurum demo that was supposed to be last month.
It's probably more likely that they get Al's original (beta) voice provider, than his second (release) one. But that's just a crackpot theory I've been thinking for a while, it's also perfectly likely they get a sound-alike instead
New livestream on Feburary 22nd!
*edit* They confirmed the delay of Aurum's demo due to technical issues, and mentioned a new feature: gender and age. Gender isn't just based on format, but a more complex form of code - I'm interested to hear it, but they won't be launch features.
*edit* They confirmed the delay of Aurum's demo due to technical issues, and mentioned a new feature: gender and age. Gender isn't just based on format, but a more complex form of code - I'm interested to hear it, but they won't be launch features.
Last edited:
Man I totally missed the stream, but I did see this: Vocal A = Arctos! I think it's a very cute design
They also posted a very short clip of a new voice, Donovan! Link to timestamp:
Ohh that's really nice! I think it's a nice gentle voice that will hopefully do well in a variety of genres
To further extrapolate on Arctos's design, since I posted it at work, I really like the style. I think it's mage-like in that the coat kind of billows out like a wizard robe and the buttons are the star shapes, but it doesn't lean into it so hard that he feels niche. It's nice!
To further extrapolate on Arctos's design, since I posted it at work, I really like the style. I think it's mage-like in that the coat kind of billows out like a wizard robe and the buttons are the star shapes, but it doesn't lean into it so hard that he feels niche. It's nice!
Audine demo!
Quite a few instances of unnatural pronunciation and odd stresses on certain phonemes, I hope this gets smoothed out closer to release - the overall engine sounds like it has potential.
Quite a few instances of unnatural pronunciation and odd stresses on certain phonemes, I hope this gets smoothed out closer to release - the overall engine sounds like it has potential.
Last edited:
It reminds me of the engine sound of Vocaloid with the nasal emphasis, which an interesting route for the engine. Audine has a very cute voice though, 
she also makes me think of Ruby?
Watching Oliver's original demo* and Her demo back to back it's wild how different they sound, but I suppose that's maybe just the nature of different producers and stages?
she also makes me think of Ruby?Watching Oliver's original demo* and Her demo back to back it's wild how different they sound, but I suppose that's maybe just the nature of different producers and stages?
Audine's tone is very nice. I like her voice a lot. I was also surprised by her fumbling of pronunciations and stresses though-- it could be a side-effect of the modifications/tools that are currently available for the backend, but it is interesting and strange. Not what I expected as far as an engine that purports to have such wide language support-- if anything, I expected Maghni to "over-correct" (or otherwise be overly precise) on pronunciations and stresses for language accuracy. Maybe the language preservation angle is less of a focus now, or the attainability of a near-universal cross-language model is being felt.
As expected and previously theorized in this thread, Maghni AI is not being released in 2024. They are retraining the base model and both Aurum and Audine.
The reasoning provided for the delay is that they want to support more languages, so everything needs to be retrained. I find this interesting. Is it because the engine lacks fundamental data for some sound, repetition, or phoneme? Are they adding a language that includes click consonants or something?
In general, I am curious why these new languages are secret when the existing (very large) language list was revealed upfront. Hype-generating mechanism??
I can only hope that this is something that they're doing because they now have the extra money and bandwidth to do so, and not because they are unsatisfied with the current results and feel they need to do something extra to "justify" the extra development time. If this were the case, that would be a dangerous prescident to set as feature creep risks introducing additional development risks and delays. Without knowing what the extra languages are, there's no way to theorize with any specificity what's going on, but again it is unsurprising that the delays have panned out in the way they have.
Hoping the next stretch of development goes well for them, curious what Audine will sound like in a before-and-after setting after everything is overhauled.
As expected and previously theorized in this thread, Maghni AI is not being released in 2024. They are retraining the base model and both Aurum and Audine.
The reasoning provided for the delay is that they want to support more languages, so everything needs to be retrained. I find this interesting. Is it because the engine lacks fundamental data for some sound, repetition, or phoneme? Are they adding a language that includes click consonants or something?
In general, I am curious why these new languages are secret when the existing (very large) language list was revealed upfront. Hype-generating mechanism??
I can only hope that this is something that they're doing because they now have the extra money and bandwidth to do so, and not because they are unsatisfied with the current results and feel they need to do something extra to "justify" the extra development time. If this were the case, that would be a dangerous prescident to set as feature creep risks introducing additional development risks and delays. Without knowing what the extra languages are, there's no way to theorize with any specificity what's going on, but again it is unsurprising that the delays have panned out in the way they have.
Hoping the next stretch of development goes well for them, curious what Audine will sound like in a before-and-after setting after everything is overhauled.
Hi, I'm Audine's developer (as of late January) and I'd love to help answer any questions about her or Maghni as a whole.
As said in the Maghni Moment, it is due to an increase in working with 3rd parties of various backgrounds behind the scenes that more languages than initially intended have been added to the base model. Due to the sheer volume of data from these new languages, retraining the base model will take longer than expected. This is the cause for the delay, no need to worry about it being the other way around! (ie it's new languages -> delay, not delay -> add new languages as compensation).
Regarding the secrecy of the new languages, some of the respective 3rd parties have requested that they be the ones to make the announcement(s). Therefore, I think that announcing them outright would maybe spoil half the surprise! :D
Thank you, I worked hard on her; her voice actress is a great singer too! She does indeed have some trip-ups, but this is largely due to the iteration of the base model in the demonstration.
I believe this has been confirmed somewhere else, however, not all of the 40 languages supported are currently within the base model; especially not the iteration that has been utilized thus far.
As said in the Maghni Moment, it is due to an increase in working with 3rd parties of various backgrounds behind the scenes that more languages than initially intended have been added to the base model. Due to the sheer volume of data from these new languages, retraining the base model will take longer than expected. This is the cause for the delay, no need to worry about it being the other way around! (ie it's new languages -> delay, not delay -> add new languages as compensation).
Regarding the secrecy of the new languages, some of the respective 3rd parties have requested that they be the ones to make the announcement(s). Therefore, I think that announcing them outright would maybe spoil half the surprise! :D