Thoughts on the Lytro Camera

So many months ago I put my name (and credit card number) down for the Lytro Camera, and just this week it arrived. Having had a few days to play with it, and having been to a Nada Surf concert, a nice restaurant dinner, and having generally just fooled around with it, I thought I would share my initial impressions.

For people who have not heard of the Lytro Camera, it is intended to "revolutionise" photography by providing a camera that records the entire lightfield that hits the lens. This should result in a camera that can take photos very quickly as it is should not be necessary for anyone (either the camera or photographer) to have to pause and find the best focus. Instead all the information is recorded, and the best focal point can be selected later as time allows. In addition to post-shot refocusing, the Lytro company also believes that future software iterations should allow some degree of retargeting, and maybe other features as well.

Quick (TLDR;) Summary

I'm sorry to say that in its current form the Lytro Camera is more a toy, than a genuinely useful tool. I had hoped that I would be able to recommend it to other's but the current version has too many issues, most significant is the final shot quality. While many of these issues lie in the software and so can be fixed, the most significant problems lie in the hardware.

In spite of that rather sad summary, I do believe the technology has huge potential and I look forward to subsequent revisions to the underlying hardware.

Hardware

The Lytro Camera itself is a beautiful piece of design, but also harbours the most significant problems. The difficulty with these hardware problems is that they are obviously not correctable without iterating the camera design, so are key to my recommendation against buying the Lytro.

Style

As I've said, the camera is good looking. A lot of this comes from the ability to remove the many nobs and buttons that are typically present on a camera, and instead use a touch screen that provides some of those controls. The only hardware controls left on the camera are the subtly recessed power and shutter buttons, and the even more subtly raised zoom control. This produces a very clean looking camera, divided into the smooth aluminium lens housing and the textured gray plastic sensor housing.

This desire for clean lines extends even to the lens caps, which is free on the usual clips, and instead is a simple square that is held to the camera by magnets.

The Good

The lens on the camera is good (at least by my not-a-real-photographer standards), as is having real optical 12x zoom (I really wish software zoom would just be left out of cameras it does nothing but harm). The camera also feels good in the hand, and zooming (once you've found the controls -- see below) was easy without moving jostling the camera.

Someone at Lytro also had the good sense to realise that when you want to take a photo there is no reason to need to hit two distinct buttons, so the camera can be powered on by the power button or the shutter button. Coupled with the very quick wakeup time for the camera, this lets you take photos almost immediately by pressing the shutter button twice. This may not seem like much, but I repeatedly found myself grateful for it and hope other cameras will pick up the feature.

The Bad

Unfortunately there's a lot more that is "bad" in the hardware than there is that is good. Part of the problem is that if you want a device to replace a regular camera you need a device that is at least as good as a camera at the things a camera can already do. There are a number

The aforementioned clean style actually makes it difficult to find the shutter and zoom controls. If it is dark (as in a concert), it is essentially impossible to find the controls without simply feeling around until you find something that is unique -- typically the zoom controls (which have a slightly different texture from the rest of the backend). But the zoom controls are touch sensitive rather than requiring physical movement of the control, so in locating the textured surface you've also possibly changed the zoom.

While using a touch screen does away with the need for many external controls, it also means that trying to control the focus target of the camera requires putting your finger in the way. This works with modern smart phones because the screen is large. The lytro camera touch screen is tiny, and so your finger is easily larger than most things that you'll want to focus on, making it very difficult to get some shots.

The actual screen portion of the touch screen is awful. It has one of the worst viewing angles of and LCD I've seen in years. Given that's your only point of shot feedback and control that's inexcusable, it's so bad that you are almost forced into being in a position where you are looking straight down the camera for any interaction. If you want to take a shot where that is not possible you get colour and gamma distortion almost immediately, and very quickly get a screen that is completely black. I found this to be a real problem when trying to get shots at both the concert and restaurant.

The actual photo quality is quite poor. The final shots are a fairly low resolution by todays standards. I realise that this is probably a result of the effective resolution being the result of dividing the actual sensor resolution by the number of focal points, but the why isn't important when just the camera as a tool. All that matters is that the final result is a very low resolution[1]. In addition the camera requires well lit environments, the low light performance is surprisingly bad, you could almost make a comparison to some of the earlier iPhone cameras. Low light shots are very noisy, and frequently even show artifacts of the microlens array used to provide multiple focal planes.

The main selling point of the Lytro Camera is that you should never need to worry about focusing a shot. It's therefore somewhat surprising that the camera seems to have a lot of difficulty focusing, sometimes his just means waiting for second or two, sometimes it just seems incapable of focusing at all without help. This seemed particularly frequent in low light, but even at the restaurant where we were well lit it had tremendous difficulty focusing on the edge of the glass. The periodic problems with focus were also matched with periodic problems with white balance and exposure time.

The Lytro also suffers from all of the problems of simplified camera interfaces -- there are just too many occasions where automatic focus, white balance, and exposure can't get what you actually want.

Finally, the lens cap. When we first looked at it Mo said "it's really clever, I wonder why no one else has done that" (paraphrased). I can now say, the reason no one else has done it, is because it makes it too easy for the lens cap to come off. I had to spend around five minutes hunting around the floor of a concert to find the lens cap after I bumped it off accidentally.

The WTF

The Lytro Camera looks dramatically different from other point-and-shoot cameras, and has a giant lens at the front. As a byproduct of this when going to the concert I was repeatedly asked not to film the concert and to demonstrate that the camera couldn't shoot film (though strangely people were allowed to take smartphones, which can all record film, into the show). Presumably the shape and big lens on the front make it look sufficiently similar to a film camera to confuse people.

Software

I included software elements of the camera itself in the previous section, but the Lytro Camera also needs desktop software as it obviously isn't recording ordinary image data. Lytro has taken what I think is a clever approach to this. Rather than including a driver or software DVD, when you first plug in the camera the system sees the camera as a drive containing the software installer. Once that software has been installed, and you've rebooted (really? a reboot in this day and age?), the system now sees it as a Lytro Camera, and runs the Lytro software when you plug it in.

In general the software is okay, it does what it has to do, it looks okay, and it is easy to get around. Unfortunately once you get past the "can I access my photos" stage, it takes a trip downhill. Happily because this is all just software most of these problems should be fixable by updates from Lytro.

To me the biggest problem is that the software doesn't support directly exporting a regular image, so your only options for sharing an image are through FaceBook and Lytro's own sharing facilities. Both require storing your images on Lytro's servers. This is frustrating because part of the Lytro sales pitch is that you can worry about focusing the image after that shot, but once you've chosen your favourite focal point your only option is to share the "living picture" that anyone can refocus. While that is fun (it really is, you can have some neat photos that way), there are also plenty of times where you get to the point where you just want to share a simple photo with the focal point you chose.

This method of sharing also has a large number of problems, the most obvious problem is that if you want to share your photo with anyone at all, it must be uploaded to Lytro's servers. Should Lytro fail, you may lose the ability to share these photos with anyone else. Every shared image also acts as an advertisement for Lytro, and I do mean this literally, every photo shared on FaceBook gets the title you specify, but instead of including your long form description it gets this:

    Lytro / <your username here>
    pictures.lytro.com
    I just posted a living picture from Lytro.com. What’s that, you ask? Click to play and find out.
				

The other, less obvious, problem with this model of sharing is that images are all display with a flash application from Lytro. This means that you cannot view Lytro images on anything without Flash (so no support on iPhone, iPad, nor the majority of other mobile devices. Win8 also apparently won't support Flash, so no viewing there either).

The obvious solution to this would be for Lytro to simply replace the current Flash based viewer with a JavaScript and html based viewer that anyone can take and embed, and use there own hosting for if they so desire.

The most significant remaining problem with the Lytro software is that the company seems wedded to the idea of selecting focal point by clicking on a 2D image. This frequently means spending a lot of time clicking in random locations in an image to try and find the right place to make the software select the focal point you want. It seems that it should not be hard to just provide a mechanism to directly select a given focal length, even if that doesn't look as cool when you're doing it (although I think it would look better). The cynical engineer in me suspects that this may be due to it exposing a finite set of focal planes rather than some kind of infinite field, which while neat probably doesn't seem as amazing.

There are also a bunch of UI quirks (a couple of distinct modes, weird navigation), that just feel annoying and counter to the polish present in the rest of the Lytro hardware and software.

Final Thoughts

The Lytro Camera is trying to bring something that is actually new to the world of cameras, and to an extent it succeeds. Unfortunately the unfixable problems in the current hardware mean I'm just not going to recommend one to anyone without copious amounts of spare cash, and even then I would still go to some lengths to point out it is still technology in its toy stage. Unlike the problems in hardware, the software can be fixed, and provided Lytro provides a way to easily share both regular and "living" pictures easily, and without requiring plugins or hosting from Lytro I wouldn't consider it a dead end. But the current state of sharing does seem fairly bad, especially given the tendency of startups with "revolutionary" technology dying/being bought and shutting down their hosting without support I am a bit concerned.

This review may come across as a bit harsh on a new company, with a new technology, but my goal has been to describe my experiences with the Lytro, in the context of current technology. That means I haven't given them any leeway just because they're new, or even when I can make a fairly confident guess as to what the reasons are for some of the short comings. Just because there is a reason for something being the way it is, does not mean that that is not the way it is, it just means you know why is that way. Hopefully future iterations of the hardware will improve upon those limitations, and Lytro will produce a camera that I would actually recommend. For now though, that isn't the case.

Related

[1] Nirav Patel reversed engineered the file format, and his output supports the idea that the sensor resolution is basically killed by the multiple focal points.