Voxta Scenes Voxta Beta Scene

Best scene to test things out and discover Voxta.

Awesome!

Awesome. And as hardware is getting better exponentially, with in a few years, we'll be able to have better and better, more realistic interactions and responses. I can't wait to see what the future holds. This is by far the best AI integration yet.

Not perfect yet. And not worth it yet from a purely consumer standpoint. But I don't regret subscribing on Patreon because nothing else comes close to this level of innovation and user choice for which AI services to use. The amount of choices might have been the reason why the reason why my initial experience wasn't great. I used NovelAI for everything except for the speech to text, and a problem I was encountering is that in the server log, every time I would give a command into my mic, it would say something like "doing implied actions" then " no implied actions" even though I clearly said "sit in the char" and it wasn't able to derive any implied actions from that sentence using NovelAI. Maybe I would have better luck with ChatGPT, but ChatGPT censors itself whereas NovelAI does not. Nevertheless, I still think this deserves a 5 star review for now. But if this issue still persists, I don't know, 3 months from now, I may lower it.

This is the best thing that happened to VAM after it has been released. Try it, and you'll see what I mean.

Unfortunately, there is no way to give Voxta 6 stars out of 5.
WARNING: It IS 'moderately technical' to get set up.
You don't need to be 'a computer person', but you definitely need to know that documentation exists AND you have to be willing to read it.

Voxta works through connecting a number of services together into one 'thing' that works really well.

But that involves you setting up and connecting:
Speech to Text (so Voxta can 'hear' you)
Text Generation (so Voxta can reply to you)
Text to Speech (so Voxta can 'talk to you')

There are multiple options for handling this locally on your PC, or in the cloud.

As with everything VaM-related, Better Hardware = Better Results.
4090 24GB + Ryzen 5900 + 128GB DRAM is *mind bending* - but there are lots of options available for those willing to power through - including running about 3/4s of the bits and pieces in the cloud (for a fairly reasonable hourly cost - f'rinstance a virtual machine with a 4090 at Runpod is 50-75 cents/hr - or about 2250 hours of Voxta-ing for the price of a 4090 of your own...)

Also:

As with everything VaM-related, it's a sandbox, and there is a significant element of 'cut-n-try engineering' your going to have to do to move beyond what's provided in the demo scene.

But if you're a tinkerer at heart, Voxta just blew the doors off this thing.

brilliant

This is a really nice scene to show off some of the capabilities of Voxta. The action inference system and also the buttons are relatively easy to understand and configure. The animations are nicely done as well. This scene is a great framework to build off of additional animations and actions, really bringing the AI characters to life.

What I'd love to see next is some more exaggerated facial expressions tied to emotions when the character is responding, maybe something in the system prompt that tells them to pick an emotion from a static list (I'm wondering if this can be done in a separate "expression" layer so as not to interrupt whatever other animation is playing). That may be challenging though, given how audio lines up with the action sometimes or how the character speaks (and every produced audio clip has some small variation too). This may be something that the community can work on though and not necessarily be tied into the beta scene.

Fantastic job guys!!!

This is an amazing ecosystem! Kudos the the creators for their hard work making such an amazing plugin/utility!