[jseliger]

In some ways, this feels like the reviews of the last couple of generations of iPhone: it can work really well in very good light and at typical web resolutions. The photo of the woman in the parka is very impressive as a technical demonstration but is also pretty blurry/smeared, to my eye. As an only camera, it's probably the best or near the best of the phones.

[zobzu]

I think the photo of the woman is terrible. I don't blame the phone for it, its very hard to achieve with small sensors, and the iphone is probably the best of the bunch. But it doesn't make the picture quality near "good-enough" for that scenario.

With slightly better lighting it can be "good enough" to be hard to distinguish from a bigger sensor camera for "web" where "web" really means ~6inch phone display - and that's great, but let's not fool ourselves ;-)

I'm also curious what the max will look like - though at the size and price it's difficult to say if buying a RX100 and a cheaper phone isn't a better option still, for many.

[Nition]

> I think the photo of the woman is terrible.

On any normal camera that photo would be a complete blur. Impossible. It's a hand-held three second exposure.

[sudosysgen]

You don't need a three second exposure with a modern camera and a fast lens. F/1.2 or f/1.4 with low-light stacking at 1/10 to 1/2 will more than do it for moonlight.

Plus, it has artifical light, a phone screen, so even less than that would be useful.

As for AF, the obvious solution is to simply have a light in the camera. Mine does, it works fine even in complete darkness. I've actually gone and took a photo of a subject lit only with a phone screen, in perfect darkness, at around 2 meters.

Result? F2.8, 1/10x3 (stacking), ISO6400 actual, ISO2066 effective.

Very, very far from a challenging situation. Actually, quite easy.

With my F/1.4 lens, a cellphone screen provides enough light to illuminate a wooden plank to a shutter speed of 1/25 at ISO4000. More than serviceable. Especially with stabilization, where I can in reality get away with 1/2.

You know, dedicated cameras also evolved. A lot.

At the issue of this expriment, I can be confident that my camera can take pictures of much higher quality in the same conditions, at a higher resultion with much less noise and more detail, and all that at half of the price. Indeed, an a6000+EF-speedbooster+Tamron28-75 2.8 EF will run you 600$ tax included, for a higher performance, and will last you a decade or more.

EDIT: for some reason, I can't answer since I'm rate limited, but I can actually do a 3 second exposure, hand held, with my camera, and get a sharper result that that. Especially if I'm tipsy. Here's how I do it - I put my camera in image stacking mode, set the shutter speed to 0.8 seconds, and fire the shutter.

My camera will then take 4 pictures with IS on for 0.8 seconds in each of them, recenter IS in the few milliseconds between them, then warp and shift the images together and stack them. Bam, 3(.2) second exposure. It's not always the sharpest on the first try, but it's a hell of a lot sharper than whatever is in the picture there. Also, the iPhone is rejecting quite a few images there, so it won't be 3 seconds and more like 1.2 seconds.

Especially at the wide focal lengths equivalent to an iPhone. It's not very hard. I was at 1/20 before to ensure maximum sharpness, but if I'm willing to either stack in post or take 2-3 shots, more than possible.

[Nition]

Unfortunately it's too late to edit my comment now but yes, I should have been clearer that what I meant was taking a hand-held three second exposure with another camera, not that that specific photo itself would be impossible.

It's a reasonable separate point to make that a "proper" camera wouldn't need such a long exposure for the same amount of light.

[sudosysgen]

Most cameras can't, I agree, but mine does. It has a 4 exposure stacking mode, but only in JPEG mode, and the stabilization is sufficient that I can get quite sharp images at 0.8x4 handled, especially at the 35-28mm equivalent focal lengths of an iPhone.

I can share some pics, if you want. It does require a steady hand.

But then again, the iPhone doesn't really do a 3 second exposure, it rejects a lot of frame so it's more like 1.5-2 seconds, which I can do reliably.

[wlesieutre]

And good luck with auto-focus

[frostburg]

IBIS and a f/0.95 lens would get you there (or a flash, which phones mostly can't synch to).

[snowwrestler]

That shot is testing the low-light autofocus with no modelling light, which I think is a test that any f/0.95 lens/body combo on the market right now would fail?

[sudosysgen]

Most modern cameras can autofocus in perfect darkness using a small AF light. Especially at f/0.95!

But really, I recreated the shot, in a perfectly dark room with a phone light as the only light source, and it focused adequately and the image came out more than adequate at 1/20x ISO 2066 @ f/2.8, or 1/30 ISO2400 f/1.4, all with good AF.

[Nition]

Ah well, F/0.95 is cheating a bit though, I meant assuming you were doing a hand-held three second exposure with another camera.

Of course it's a reasonable point you make that a "proper" camera wouldn't need such a long exposure for the same amount of light.

[inasio]

This is on a handheld three second exposure in the dark. Not sure you would do much better with an SLR. I have on Olympus OMD EM5 with great OIS and that shot would be pretty challenging

[spike021]

It was a 3-second hand-held long exposure in apparently little to no light. I'd say the fact that her features are somewhat defined is pretty decent.

It seems the point of it in the first place was to see how well the night mode works in near pitch-darkness for portraits.

[KineticLensman]

> I think the photo of the woman is terrible

This is partly because the flash has blasted her face with light losing all the shape except for an ugly shadow under her nose. With my Nikon I could try at (very) high ISO for a natural-light shot or use an off-camera flash to shape her face in a more flattering way. But perhaps people used to the very flat / boring light common in mirror selfies won't care about this.

The blur on the edges of the fur look digitally smeared rather than optically blurred.

[edit - okay, I didn't spot that he used off screen light rather than flash. But the effect is similar and hasn't given any pleasing sculpting of the subject. Other processing may have made things worse]

[kristofferR]

He didn't use a flash for that image, the only light was from another phone screen. And it was a three second hand-held, it's impressively clear for what it is.

> This Portrait mode w/ Night mode 3-second exposure was shot with the iPhone 12 Pro in my right hand. Simultaneously, in my left hand, I held my iPhone 11 Pro and used the light emitting from the screen (not the flash) as an off-camera fill light for Esther’s face.

[inasio]

I don't think there was flash involved in the shot, it says he held another phone on his left hand and used the light from the screen (not flash) as an indirect light source. The washed out look I think is from software averaging to correct motion blur from the 3 second exposure.

[sudosysgen]

The blur is actually insane amounts of noise reduction, it's not an attempt at bokeh, I think! You can tell because all other shadows are just as blurred out.

I think that it doesn't look flat just because of the flash, but really as a mix of incredibly severe noise reduction as well as a ton of dynamic compression necessary to keep the background in roughly the same exposure.

[wlesieutre]

He says it’s “Portrait mode”, so it does have the fake bokeh on the background in addition to any smoothing from night mode denoising. That’s why he also mentions the depth map for the fuzzy hood.

[bitL]

It's unrealistic to get off-camera flash on iPhone unless two could be paired together with a friend holding iPhone acting as a flash. But what could be done easily is a ring flash around the rear camera that would make most portraits bearable instead of horrible basic flash.

There are also Deep Learning models that change angle of light on a final photo, so you would be able to get your Rembrandt lighting that way soon.

[ssijak]

What percentage of people would rate portrait photo with bokeh much much better from a DSLR than from an iPhone 12 pro (or any other top Android phone) in a blind test? I would say that more people will prefer the phone photo, unless you edit the dslr photo (which phone is doing automatically unless you shoot RAW on iPhone).

[yladiz]

I don't know how you would define "better", but if you mean higher quality, I'm certain that most people would be able to pick out the DSLR version in a "blind" test, even if you used the same approx. lens focal length on the DSLR and the phone (e.g. 52mm when comparing to iPhone). I would guess people would prefer the DSLR one, but it's simply a guess (and I'm biased because I know more or less what to look for).

[asdfasgasdgasdg]

I'm not sure if I'd say that most people would be able to tell. But I'm sure photography people would be able to tell. Unless you're a wannabe influencer or a working professional, what matters is not whether other people would be able to tell. Nobody cares about your pictures. What matters is if you can tell, and how it makes you feel.

[jonplackett]

I think if you let them zoom in to the edges they'd pick it easily enough. My iPhone always blurs a few little details, little part of a hat sticking out, thinking it's the background.

[zobzu]

the picture we're talking about is taken at night, and looks like a painting, this has nothing to do with bokeh or the need to edit a picture taken with a large sensor mirrorless camera

[ValentineC]

> What percentage of people would rate portrait photo with bokeh much much better from a DSLR than from an iPhone 12 pro (or any other top Android phone) in a blind test?

As an iPhone 11 Pro + Halide user myself, I daresay a DSLR photo with a decent portrait lens would definitely look better — or at the very least — more natural, than any iPhone photo that relies on "computational photography" fake-bokeh.

[numpad0]

There are surprisingly little “cultural” differences between iPhone and dSLR, iPhone shots look like a Nikon at wrong ISO.

[grecy]

I would guess close to 0% of people, IFF we're looking at the photos side by side on a phone screen, or maybe a crappy laptop screen.

Put them both on a hi-res display, or HD TV or (even better) print them out in large format. Then there's no comparison.

So it really comes down to how is the photo going to be used.

And for 99% of photos taken these days, viewing it on a small screen is the purpose, so the phone camera is perfectly good enough.

[sudosysgen]

For this picture specifically, which is the one OP was talking about: https://images.squarespace-cdn.com/content/v1/50730c37e4b03a...

Yes, anyone could tell the difference between this taken on a dedicated camera and this taken on iPhone. Really, anyone. No matter the resolution.

It really does just look like a blurry mess. A cheap 150$ DSLR with a bottom-bargain Chinese or very old lens (at say, F1.4 or lower) would provide significantly better results.

[ssijak]

I think that this was the image referenced with the "woman in a park" in the initial post https://images.squarespace-cdn.com/content/v1/50730c37e4b03a...

[sudosysgen]

Oh, I read "woman in a parka" instead, and in combination with "blurry, smeared" I concluded that it was the image at night.

[TrainedMonkey]

You are correct that a larger sensor would capture more photons, however I think you are not comparing apples to apples. That picture is a snapshot with basically 0 visible light and was made possible by capturing a lot of channels and running a remarkable software enhancement. AFAIK, there is no API to do anything like that with uploaded raw picture (0). So to get a similar quality picture with a larger sensor you will need a long exposure time... and an average human subject shivering in a cold environment will come out way blurrier.

Good portraits are usually made by excellent lightning conditions. As a portrait this picture is average, but when you take conditions in the account the result is quite amazing.

(0) That is an assumption I am making which might be wrong. You could also definitely do this by hand by investing 100-1000x more time.

[simonh]

As a photo for a magazine or other professional use sure. As a photo of your daughter the first time you've managed to get together after the lockdown late at night after a long journey, effing incredible.