In my roundup of what I saw as the 11 key AI features announced by Apple at this year’s WWDC, I said that the camera features seemed extremely well thought through.
In the two weeks since, my view has been very much reinforced. The Clean Up tool is now significantly more powerful, while the Reframe and Extend tools address similar real-life needs …
I got my first camera as a birthday present at the age of 14. It was a Zenit-E, a Soviet-era Russian hunk of metal that weighed about as much as a car and had exactly zero electronics or automatic features.
Technically, it had a built-in light meter, but this was a Selenium one that could just about measure the difference between a pitch black room and being pointed directly at the sun. An external light meter was needed, and every exposure had to be manually configured. It was ugly, heavy, and clunky – and I absolutely adored it.
If you’d told me then that I would one day be using a phone as my primary camera, and getting far higher quality results than even much later SLRs, I would have enquired as to your use of recreational chemicals. And yet here we are.
It would have seemed just as unimaginable a decade ago to talk about DSLRs being discontinued, even in the pro market. And yet Nikon has ceased manufacturing them altogether, and while Canon keeps making existing models, it hasn’t developed a new one in years and likely never will again. Mirrorless cameras dominate the market, with industry surveys showing this to be true even among professionals.
We may not yet see many professional photographers routinely using iPhones for their shoots, but it’s certainly common for them to be praised for their capabilities. Today’s iPhones are pretty serious cameras.
Of course, there’s only so much you can do with tiny sensors and plastic lenses. You don’t, for example, get very much control over depth of field.
Apple was one of several companies to realise that clever photo processing could emulate results previously only possible with higher-end cameras. Portrait mode provided a means of mimicking the effects of a shallow depth of field, for example. Early results weren’t great, but the feature has become increasingly convincing over the years.
Today, a great deal of computation goes into the processing of iPhone photos quietly in the background.
Apple also started providing the means to correct mistakes after taking a photo. For example, if you use Portrait mode at f/2.8 but later decide you want more of the shot in focus, you can retroactively change the simulated aperture. If you focus on the wrong person or object in Cinematic video mode, you can move the focal point during editing – either on the iPhone or on the Mac.
If you framed a photo poorly such that there was an unwanted person or object in the background, the Clean Up tool allowed you to remove it afterwards.
In iOS 27, Apple has made the tool Clean Up tool significantly more powerful, and supplemented it with two additional corrective features: Reframe and Expand.
Reframe allows you to artificially shift the position of the photographer left, right, up or down and see the photo recomposed as if it had been taken from this angle.
Expand allows you to change the aspect ratio even where this requires adding content to the frame that wasn’t present when the photo was taken.
What impressed me about Apple’s choice of AI tools is how they address real-life needs. The company isn’t using AI to do clever things just because they are clever, but providing features which are genuinely practical and useful.
The Clean Up tool now delivers better results when used in a similar way to previous versions of iOS. For example, this photo has a somewhat distracting ferry to Cher’s left:
So let’s remove that:
There’s that ferry again, and also a mooring buoy to the left.
Let’s remove both.
But let’s give the tool a real challenge. Here we have two unwanted people in the shot of Ambara, especially given their high-vis clothing.
A few seconds later, it’s as if they were never there.
Any photographer knows that a very slight shift in angle can completely change the composition. Here, Ambara is nicely centred in the arch, but at the cost of including an ugly recovery truck on the left and a lot of cluttered traffic on the right.
It might have been better to sacrifice the arch symmetry in order to position myself a little further to the left. Reframe allows us to do exactly that.
We have a similar situation here with some distracting people in the background to the left.
A very slight shift to the right would hide them from view, albeit at the cost of losing separation between Cher’s hair and the tree.
(We’d ideally want to remove the piece of litter also, but we can’t currently combine Reframe and Clean Up.)
Again, let’s make things more challenging. This composition looks slightly uncomfortable because it cuts Ambara’s arm.
A shift in angle to the right would correct that, but this is a pretty big ask of the reframing tool as it essentially has to invent much of her arm. And yet …
Sometimes the tiniest of changes can make a difference. For example, in this shot, I would like just a little more separation between Cher’s windblown hair and the church spire in the background. The tiniest of reframes provided the solution.
It’s exactly the same here, where a microscopic change of angle provides separation between Ambara’s elbow and the side of the arch.
Finally, the Extend tool allows us to include elements that were outside of the frame. This is even more challenging as it requires the iPhone to make up plausible and realistic content. Here we can see a couple of examples with London’s Walkie Talkie building.
In the first one, I’ve cut the building vertically. The extend tool has done a convincing job of recovering the lost edge of a building, even if it does have to make up a building in the background to do so.
In a second example, I’ve cut it horizontally with the top of the building lost in the same way. What’s fun here is that the AI has seen all of the cranes in the background and concluded that there is likely one on the fictitious top of the building behind it. If you didn’t know the scene, I think this would look perfectly plausible.
Of course, this is the first developer beta, and there are the inevitable glitches and crashes, but I think Apple is on exactly the right track with this kind of thoughtful and genuinely intelligent use of AI. I really can’t wait to see what else will follow.
What are your own thoughts about this development? Please share in the comments.
FTC: We use income earning auto affiliate links. More.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。