Ellis Rovin

Can you be like, make this a picture of the back of them?

So Photoshop can currently do that.

That's a feature that just came out in Photoshop.

You can literally rotate any photo and it'll just generate what it should look like.

It's a people though?

I thought it was only objects.

I could be wrong.

No, people too.

So my question for this isn't like

because you're taking a spatial photo, it's using the wide sensor and the main sensor.

Can you just change the perspective between that little parallax that you get between the wide and the main?

Or is it like you can rotate it like fully?

That was sort of my question about all of this is like,

how how much are they just going to like you know diffusion transformer model their way to like new information and how much are they going to be capturing like quote-unquote metadata with the other two cameras or the other camera and then inputting that into a right probably a little bit of both if you always have the ultra wide frame going then you can always have extra information about what's around what you take a picture of with the primary camera right so maybe if they're smartly always capturing some of that information around the outside edge and then

running it through a model and including that in a slight reframe i could see that not being insane and that would allow the generative expand to also use other things that were actually there right because the isps in these in these smartphones they have like two or three now and they're they can just run all of the cameras all the time yeah so maybe they're just sampling from every sensor and then if you use generative expand it it stores that metadata like in a compressed format it

I don't know if I'm wrong, but the iPhone camera system, excluding the single camera ones, kind of already do this, right?

Like every photo you take, there's image data coming from the other lenses too.

Google first did that with like Super Res Zoom.

Yeah.

And I think Apple does that now as well.