Widescreen in-fill (generative AI)

More important than B&W, 3D or HDR conversion, by far I think, is: Transforming 4:3 into 16:9

Or any other aspect ratio conversion — where cropping isn’t an option.
The “clever” software needs to see the panning & tilting in the video, and use that for hints.

Strong hints!
This is even more appropriate for video AI than photo AI — there’s actual data in neighbouring frames.

I’m thinking about this after seeing Adobe’s new Generative AI features for Photoshop.
Any technology where you “fill in the blanks” and overcome occlusion … is going to help here.

IMO: I think there’s far more scope for WS conversion than 3D or B&W, maybe even HDR.
The future of SDR looks fairly healthy for viewing on TVs … maybe PCs will auto-upmap one day?

For some movies this could work super well. Others not at all.
It’s a good idea for sure.

Yeah — you’d need to pick & choose scene by scene.

But that sort of editing is easy (and fun). There’s also 50% of scenes that work with cropping.

By combining cropping, semi-cropping and AI WS, you could complete anything nicely.

Well dang. I’m surprised it doesn’t have this already. I have Video AI and Gigapixel AI, and only just now needed to do some 4:3 → 16:9 infilling, but I see it’s not there. Yet, there are other video tools coming out (or are out) with infill. Would have thought Topaz, with a premium, would be on the forefront of this!!

1 Like