Week 469 was posted by Charanjit Chana on 2026-10-20.
I only have a basic example to share here as I haven't found a photo where I'd want to use it for real, but the capability is very impressive and it's something I've played with on a few photos and wanted to share the outcome of one.
Tested with iPhone and DSLR photos, they all work and they do adjust the composition of a photo. We saw it demoed at WWDC '26 but I didn't even mention it back then because I knew I'd not be upgrading this year and wouldn't have access to the feature. Then I realised a week ago that I should have access in Photos for on my Mac.
Taking a recent selfie of me and the kids, I was able to subtly recompose without much effort. Push it further and it takes cues and starts generating extended portions of the photo. I wouldn't say it's obvious but look closely and you can start to get a sense for what might be AI generated.
I did try with holiday photos but to me they begin to look artificial. I also tried with Live Photos that I'd set to Long Exposure. It worked, but it then exaggerated the parts of the image I was trying to get the effects on and removed the ability to watch the Live Photo.
So here's an image I tried it on, taken with a DSLR.
The top image is the original, edited in Photos. The second and third are the same image but using the Reframe feature. The middle one pushed the limits of what was in the photo. The rear of the pumpkin would go missing if I turned it all too much. The third used zooming, rotating and reframe all together to achieve a different angle on the photo and you can see less of the top of the pumpkin.
Is anyone else using Reframe meaningfully? I can see it could be really useful for photographing kids who are unlikely to sit still but beyond that I don't know when I'd use it. It is really impressive though.
I also wonder if tagging people in photos correctly helps the algorithm to better position people you take photos of regularly?
Tags: