I am glad they have stuck to their "community update" schedule thus far. Similar to what was expressed earlier, I'm not sure how much of it was highly enlightening-- good for increased clarity, more information, but also, what it is, is effectively more promises being made and more quantitative statements. So rather than being "progress report" they are effectively additional marketing/hype posts. There's nothing wrong with that, but hopefully before release, they establish very clearly what Maghni AI at launch actually is. What has changed from the crowdfunding round... what is the same... what is being delivered on... what is not.
Her new design is very very cute, I like it. Demo-wise, I think I have the sense that maybe the voice shown here is more stable than past demos, but the tone is thin, and the pronunciation still muddy. Will be interesting to see if the engine developers call this out as something using their "new retrain", or not.
I don't expect Maghni AI to sound like SynthesizerV, and it shouldn't, it should be its own thing if it wants to compete (maybe not compete-- but coexist on the market). But if the engine's lofty promises of novel language implementation and accent control do not materialize, and the voices themselves remain muddy, then maybe the user experience will be what they hope carries it over the finish line in users' hearts?