Still sounds AI, you can tell they exaggerate all the tone and trailing "high scoring expressive sounds" like your job depends on it.

Yes, I find nearly every "SOTA" voice model I try intolerable to listen to because of the fake exaggerated expression/emotion. It's actively distracting because it pulls focus to emphasize randomly. ChatGPT Voice models are so insufferable to put up with for a conversation longer than 45 seconds.

All I want is a clear, technically flawless, even/restrained "computer voice" for pretty much every use case (except audiobooks). But that doesn't make for splashy demos/score well for RLHF raters.

Yes, I wish there was more focus on correct pronunciation over emotion. You need a mechanism to control/guide the voice which gets under balance right between not having to specify everything and still letting you fix certain cases (where you know a certain sense of a word is meant)

Sure a fast response on HN would make people switch. Try better rates, infra, limits etc.

The new default is 6.0 Sol high. Escalate to Astra-medium. If usage is tight go luna6.0-max

their pipeline would be to just prompt it's internal next gen models to create a jev copy given all the data they have as a first pass.

This is the real moat, the training data, they even said it, it's the meticulously crafted data that they bet on

Even then, TypeSafe AI point out that "Jev doesn't have deep knowledge of niche domains, but you can supply context to help it decide. If you’d like Jev trained on your use cases, let us know."

This is same case with LLM's which allows you to also fine tune it.

Not really. SOTA LLMs have much larger context windows.

but the author has Lara.

in two weeks, a chinese lab will have a Pev-2.7-flash-qwen for 0.00004cents/million

if is closed source, how does he know there isn't some breakthrough he doesn't know?

If it's closed source, how do you know it's not technology given by aliens from the 43rd dimension running on quantum computers enabled by discovering that P=NP and finding a linear time reduction from NP to P?

Occam's razor is that it's probably not all that different unless there is some specific reason to believe otherwise.

> "reviews"

Do you really think we can sign off on a 100 billion-element analog circuit gifted to us by a malicious adversary?

We can't even keep our own CPU's reliably free of security exploits (Spectre/Meltdown and family); and the only "adversary" there is plain bad luck. Not an active adversary. Yet, all the engineers at Intel/AMD put together couldn't uncover those things before launch.

I emphasize analog because there's classes of circuit bugs (like Rowhammer) where the digital net is correct, and it's weird physics in the analog world that allows privilege exploits, by actors who know where the analog assumptions break down. There was a researcher a few years back—I wish I remembered who it was, there was an HN thread—that demo'd a simple digital circuit with analog gadgets that completely changed what the circuit did, and which were so insidious no human would ever find them.

Just like all the current cohort of software engineers will review all of the vibe-coded slop they shovel into their releases. Right……