What to buy, and how to get the most from it.

An AI camera failed to capture the magic of CES

Memories made, not saved.

Sponsored Links

An AI camera failed to capture the magic of CES

Relonch wanted me to fall in love with photography again this CES. But its camera is so radically different from everything I've used before that I struggled to put my faith in its promise.

The company is based in Palo Alto, California, and its pitch is simple, if very Silicon Valley: a camera as a service. You hand in your old shooter (yes, really) and in return you get the 291, a unique, leather-bound DLSR-shaped camera. It has an APS-C sensor; a fixed, 45mm-equivalent lens; an electronic viewfinder; a shutter key; and, importantly, a 4G radio inside.

The 291 uses that radio to send raw files to Relonch's servers. Once they're there, AI scans through your shots and picks the best ones. To do this, it identifies the individual elements in the photos using computer vision and judges your composition. It'll then process the raws, individually lighting and coloring elements before applying its own crop and sending them back as JPEGs.

You receive a batch of photos each morning, which is key to Relonch's business model. The idea is you choose the photos you love as part of your morning ritual, which reminds you to take your camera out again and keep snapping.

The 291 itself is free. The photos are sent to you as small, watermarked files, and you have the option to keep them, which grants access to the full-size file (as large as 20 megapixels, depending on how the AI has decided to crop it). Each photo you keep costs $1, and you start your account with the market value of the camera you handed in as credit. Oh, and if you decide you want to pick a photo at a later date, you can always go back and buy it. Likewise, if you don't like the 291, you can hand it back in exchange for your old camera.

That financial proposition is what intrigues me most. Over the past five years I've spent $3,000 or so on various cameras and lenses. I've probably processed and kept maybe 300 photos, outside of work. (Of course, there are another 30,000 or so that are gathering digital dust on various SD cards and hard drives.)

I picked up my 291 the day before CES was set to begin, briefly meeting with Relonch co-founder Yuriy Motin, who explained the process. After, I headed out for a meal with a few colleagues at a pretty bad Las Vegas eatery and snapped away.

Using the 291 might feel natural to those used to a point-and-shoot or smartphone camera, but for someone accustomed to aperture rings and settings dials, it was disorienting. There's no way to zoom or manually focus and no way to check on ISO or aperture. This is intentional, of course, as Relonch believes its AI can balance the photos and doesn't want people distracted by such things. When coupled with the lack of screen, though, it introduces an uncertainty to proceedings.

The following morning, I received 15 photos via the Relonch site and went through them one by one, deciding (in Relonch's parlance) which were Remarkable and Not Remarkable. I ended with eight, at a value of $8. Were I not using the 291 for work, probably only a couple of them would have made the grade, but I was conscious of my need to have a decent bank of photos by the end of the week.

The photos were generally OK. I was impressed with the way the AI processed a chance meeting of Engadget and Verge editors, which was taken in darkness but exposed well. Yes, it's noisy and it's not going to win any awards, but it was a photo I wanted (I've worked for both sites) and I'm glad it panned out that way. Others taken at and around the restaurant were generally good, although I didn't appreciate the weird square crop on the photo of Nate (Ingraham) and Jess (Conditt).

My first day with the Relonch 291 (the crops were picked by the AI).

My plan was to continue to shoot for a couple more days and document the results. Like all best-laid plans it quickly went awry, but that wasn't entirely Relonch's fault.

I spent most of day two in our trailer, frantically writing up an article with a tight deadline. As such, I'd only taken a couple of photos by 3 PM. As I jumped on a bus to Faraday Future's FF91 launch event, that thought was weighing heavy on my mind. So much so that I ended up snapping a shot of the Vegas sunset out of the bus' dirty, UV-filtered window, which is not something I'd usually do. (I am very much about portraits rather than landscapes.)

By the time I reached the venue (which was held in a tent across town from our CES trailer), the sun had set, and I was in panic mode. I quickly grabbed an exterior shot and headed inside to find interesting subjects while still attempting to do my regular job.

The AI processing isn't consistent.

Ironically, my job that night was to support our auto expert Roberto Baldwin by taking photos of Faraday's new car, but I couldn't use the 291 because I had to publish them the same day. Not for the last time this week, I had to bring a second camera along. After I was finished with work shooting, I took a few photos of some prototypes, a couple of my colleague Robbie and then some of the maelstrom that followed the event.

The next morning, I received almost nothing but disappointment: There were precious few photos worth keeping from the event. Some of that can be blamed on lighting conditions or a last-second shake of the hand, but I'd captured some nice images at the event for Engadget and was surprised at the lack of quality photos from the Relonch.

A lot of the images were just ... unnatural looking. Like someone had used an aggressive Instagram filter or a cheap "HDR" application. I chose to keep two photos of Robbie (above), to highlight the difference between the AI's processing choices. Other images were better, and the sunset shot actually came out OK, if a little too saturated for my tastes.

On day two, I found the AI processing overaggressive.

On my final day with the 291, CES struck again. I attended a briefing with Razer for the Ariana projector and Valerie laptop, which I didn't want to take the camera to. The meeting was under a timed nondisclosure agreement, and, although I'm probably overcautious, I was worried about randomly sending photos of secret prototypes across the internet. After the briefing, I sat inside the trailer processing the images I took for work and writing up other articles.

Then, 3 PM came again. I'd taken no photos with the 291 and needed to rush across town to an ASUS media suite, where I'd be taking photos of the company's new Zenfone AR. Again, I brought my personal camera as I'd need the images quickly, but once I was done with work, I hung around the periphery for 10 minutes, waiting for the video shoot to wrap. I used the opportunity to take a few snaps with the 291 before traveling to an NVIDIA keynote that involved writing and taking photos of NVIDIA things and then to a live show where I met with people and also didn't take any photos. I woke the next day to four thoroughly unremarkable shots in my Relonch photo box.

I'm not a professional photographer but I do take a lot of photographs, and I'm confident enough that it's the camera, not me, at fault here. At the ASUS Zenfone AR shoot, lighting conditions were fantastic. On my personal camera, I was typically shooting at 1/160th of a second with ISO set to 640 and f/2.0 aperture. Why, or how, had the 291 failed under these conditions? I'm not entirely sure.

The photos I received from Relonch were all either fuzzy or blurred. They weren't massively noisy, but they also weren't anywhere near sharp. Relonch photos don't have any metadata, so I can't tell what the settings were, but for an almost stationary subject to be blurred, it must have been shooting at a very slow shutter speed, which makes absolutely no sense given the lighting in the space.

Day three was marred by confusingly poor photos.

Relonch's business plan makes a lot of sense to me. I love how passionate the company's founders are about photography, and I love the concept of an AI delivering me perfectly edited photographs every morning. I love the pitch that my life might be as photogenic as anyone else's, if I only took my camera to capture it. But I don't love the 291.

The 291 feels nice. It looks striking. But inside its leather-bound case lies a very average camera. The hardware package feels equivalent to entry-level DSLRs but without Nikon's or Canon's optical know-how and handling or the ability to switch lenses. To make matters worse, its software is clearly imperfect. The camera doesn't appear to be making the right decisions with regards to shutter speed, focus and aperture, and the helpful AI autoprocessing is inconsistent and sometimes too aggressive.

Part of Relonch's pitch is that you always have your camera with you, capturing everyday moments. When I collected the 291 from him, Relonch's Motin pointed to White House photographer Pete Souza, noting that the best shots of Barack Obama are the candid ones taken between big events. My CES was full of such moments. I met with colleagues and old friends, people who I see at most once a year. I made new friends and had memorable conversations. I watched two colleagues embody Elsa, if only for five minutes. Between the stress and exhaustion, I laughed, a lot. But after a couple days, I didn't trust the 291 to capture those moments.

A camera as a service remains a truly interesting proposition, though. It was probably unreasonable to hope that Relonch had nailed it on its first attempt, just as it was unreasonable to pick my busiest week of the year to test the 291. Assuming it has the funding to continue pursuing its dream, Relonch still has my attention. Its AI will only get better, as, I'm sure, will the 291. Early testers are apparently happy with the 291, and Relonch says its algorithms will improve, learning from everyone who uses its cameras. The 291 is in a limited launch, and Relonch is currently only accepting photographers into the program who meet its criteria.

I will definitely try one of the company's cameras again, this time carrying it around my home city, without the pressure of a trade show, and I'm truly excited to see if the company can make this business model work. I'd also consider just using the AI processing, if the company is able to perfect it and offer that as a stand-alone service. (It's not in the company's immediate plans.) So many of the photos I take are left untouched, taking up needless space on my laptop, purely because I don't have the time or inclination to process the raw files. Relonch could be onto something here, but it has to improve its software to have a hope at succeeding.

Click here to catch up on the latest news from CES 2017.

All products recommended by Engadget are selected by our editorial team, independent of our parent company. Some of our stories include affiliate links. If you buy something through one of these links, we may earn an affiliate commission.
Popular on Engadget