Logan Kilpatrick

There are cases where I just wanna type and I don't wanna speak because I have been speaking too much or I'm in an environment where talking out loud would be weird.

I feel that.

Yeah.

I think you'll sort of see this ubiquitously across all the products.

Like we actually have this in AISGDO sort of to a certain extent already.

We have this like yap to app thing that I think I've shown you all before.

And I love it.

It's nice because like sometimes, especially when I'm,

According to your point, not being deeply thoughtful and I'm stream of consciousness coming up with my idea, I just oftentimes verbally say what I want and just enumerate all the things and then let the model do the dirty work of coming up with

a well-formulated prompt based on my... Make that a plan.

Yeah.

64 seconds.

Now the barrier is doing the TTS and generating all the other stuff.

So you should do these generations offline from a TTS perspective so that your users don't have to wait a bunch of time in order to do that.

TPUs will help.

More TPUs is always part of the solution.

One of the interesting things is, and we're working on this for the audio stuff right now, and actually this is the difference between the live audio model and the TTS model.

The TTS model, at least in this example, is not streaming

the stuff back it's like doing a full generation and then it's like giving me the fully generated stuff so i think you can actually like kind of bridge this like hybrid real time text to speech not text to speech um or like asynchronous synchronous use case by just like streaming back the audio tokens um

so i think we actually are going to do that which will make like this example that i'm trying to build here like a lot faster because it'll just work out of the box um so i'm excited for that to be honest i think it'll be it'll be awesome to not have to wait to generate stuff logan before we let you go i know you've got other appointments and a busy day ahead of you and hopefully some time off in the near future uh