WEBVTT

00:01.000 --> 00:02.920
So what is Google Vo3?

00:02.960 --> 00:05.320
It's the state of the art video model by Google.

00:05.600 --> 00:11.000
The first model that's actually usable, in my opinion, and there have been a lot of great video models

00:11.000 --> 00:15.560
up until now, and they've been getting better really rapidly, but this feels like a real step change.

00:16.160 --> 00:21.160
It is still a diffusion model so similar to stable diffusion like the original.

00:21.560 --> 00:30.080
Obviously, with several generations of improvements and the ability to generate 3D instead of 2D images.

00:30.400 --> 00:36.200
So the really interesting thing is that it takes the full motion into account and kind of predicts where,

00:36.520 --> 00:38.120
you know, the next frame is going to come from.

00:38.120 --> 00:41.520
So it keeps it stable across that whole, that whole shot.

00:42.640 --> 00:44.400
So it was created by Google.

00:44.400 --> 00:49.800
And you can access it via their labs product, Google Flow, or via the API.

00:49.840 --> 00:53.680
It's also in AI studio, which is typically where I use it.

00:53.720 --> 00:57.520
I'm currently using it in flow, which is what that screenshot is here.

00:58.160 --> 01:03.600
Just because you get a lot more credits and there's some cool features there, so you're not paying

01:03.640 --> 01:06.360
per API call, you get like 25,000 credits.

01:07.720 --> 01:10.120
So what are we using it for?

01:10.120 --> 01:10.760
What's good about it?

01:10.800 --> 01:12.720
Well, one is just the video is a lot more.