The Lytro Illum camera pioneered light-field photography.
"Instead of capturing it on a single plane, freezing an image in time and space, a light-field camera also captures the direction in which light was moving. Its processor then essentially renders a 3D scene, complete with the knowledge of distance between objects. A light-field photo represents not only everything in the scene, but a spatial understanding of the things in it." [1]
It's amazing from a technical perspective, but the image resolution was low (only 4 MP), and the price was high ($1600). [2]
I'd hoped that the technology would help the camera market find a new selling point to compete with phones, but apparently that hasn't happened. It's a pity that Lytro is disbanding instead of selling off to Canon or Nikon.
Pretty wild to click through to the Verge review and all the photo embeds are dead because they were using Lytro's services. Media is so short-lived these days.
Media tries -- and is able -- to do a lot more than it did in the past. e.g. as terrible and annoying as linkrot is, most webpages are able to reach an audience far larger and far more disperse than what could be served by any library or printing press. Nevermind having the passive/latent feature of being cheaply and (hopefully) perpetually archived by services like Internet Archive.
The Lytro embeds, unlike physical media (negatives/photos that get lost) could still be revived if Lytro would open source the code. The existing embeds might be dysfunctional, but I would hope the media data could be mirrored (from the Lytro servers) and played via embeds hosted on any server. I'm guessing Lytro photos never reached critical mass for people to really call for this archival effort.
> most webpages are able to reach an audience far larger and far more disperse than what could be served by any library or printing press
Most webpages don’t have the reach of a mid-sized city newspaper in the 20th century. A few approach the reach of a major city newspaper, wire service like AP or Reuters, or radio/TV network – think about what percentage of the world could hear the BBC! Facebook or Google have those numbers but the vast majority of websites have far fewer visitors and less revenue per reader.
What’s different is that it’s much cheaper. You couldn’t run a printing faculty with a couple of people and laptops but a ton of websites do pull that off.
> The Lytro embeds, unlike physical media (negatives/photos that get lost) could still be revived
This leaves out a lot: if you have physical media, the risk is distributed and has been very well understood for centuries. If my house burns down, you don’t lose your family papers & photos. Important works are widely distributed and there’s rarely a single event which destroys all of them – even things like the Cultural Revolution didn’t manage 100% suppression.
Digital works are easier to make exact copies of but generally worse on every other aspect. In the case of a hosted service like Lytro, the complex proprietary stack is a huge risk but so is something as simple as archiving all of the data – if they’re doing any sort of on-demand loading you run the risk of only downloading the fraction of the data which displays initially without user interaction, and if they’re doing any sort of browser detection or rate-limiting you also have to be careful to deal with those. The web archiving community has years of experience with these kind of issues but it’s expensive and makes it hard for most people to archive their personal data.
Most webpages don’t have the reach of a mid-sized city newspaper in the 20th century. A few approach the reach of a major city newspaper, wire service like AP or Reuters..,
If you look at the 1990 circulation numbers, they are small for mid-sized cities, and even for large cities. The NY Times circulation was just over a million, there are hundreds of webpages with that kind of reach now.
They take a little while to load once you click on them; it's dozens and dozens of files.
The data for the embeds may still exist on Lytro's CDN, if you know the UUID for the image, and could be rescued and rehosted. After they shut their hosting service down last year, I crawled every embed the Internet Archive had, and downloaded the additional JSON and JPEG assets necessary to restore the embeds. https://www.archiveteam.org/index.php?title=Lytro
I just tried the first batch as a torrent and it worked fine for me. Note that IA torrents are hosted as web seeds, so if your client doesn't support that, you won't have any seeds.
Also note that I don't think the hosted versions can be imported back into the Lytro Desktop software, if that's what you were hoping for; they're just stacks of JPEGs, JSON metadata, and the depth map data.
Rehosting an individual embed should be pretty straightforward, just replacing the CDN URLs with relative or absolute ones to wherever it's been rehosted (or to Wayback Machine URLs). I'll be writing up some details at some point, but don't wait on me.
Rehosting an entire user's "pictures.lytro.com" gallery was trickier than that, I don't think I was able to get it working locally.
- the camera itself crashes about once every couple of minutes, and that is not an exaggeration. You literally need to keep putting the battery out and back in, and that is not isolated: every reviewer also mentions the same.
- the software generates a depth map from the light-field image. This depth map is often generated wrong, with blotches of foreground content being treated as the background, and vice versa. You have to manually export the depthmap as a grayscale image, fix it up in Photoshop, and import it back.
- when you share the image on lytro.com, it compresses this image and breaks it up into a few layers (presumably for the JS viewer), which significantly degrades the quality of the image
- the perspective effect isn't real, and uses content aware fill. This means that there are serious visual artefacts with this feature.
It’s not _exactly_ the same as how the Lytro worked, but you can achieve a small degree of fixing focus in post with the data captured on any of Canon’s newish “Dual Pixel AF” sensors that are increasingly found in their DSLRs. The technique is not a million miles away from what lytro did, in that there are effectively two photosites at each pixel measuring the light hitting it at slightly different angles, which is what gives the ability to fix minor focus issues in Canon’s own RAW processor. The feature isn’t supported that widely in third party RAW converters yet though.
This isn’t really all that comparable to the Lytro - this is effectively a hack that takes advantage of some Panasonic camera’s ability to shoot 4K jpeg bursts at 30FPS - it racks the focus as it shoots a burst so you can pick from a range of focus points. The subject can still move noticeably in between shots, it’s not a single shot style solution like the Lytro has. I have a Panasonic m43 camera with this feature and almost never use it, It’s just a fast implementation of focus bracketing and you take a pretty big resolution/quality hit.
The new “Dual Pixel AF” sensors in some newer Canon DSLRs is much more like the Lytro, in that focus can be adjusted slightly in a single image capture, no burst capture trickery.
Yes, it's just focus-bracketing and that's as close as any non-Lytro cameras come to the Lytro feature as far as I'm aware
Reading a description of the Canon Dual-Pixel sensor feature https://www.canon.co.uk/cameras/eos-5d-mark-iv/dual-pixel-ra... ...it does not seem that it allows to adjust focus after the fact. It's something different again from Lytro and focus-bracketing.
As far as I can tell what it does is to capture depth information into the RAW file. I don't see how it could allow refocusing, but the examples show it allows you to selectively mask areas of the image based on distance from camera.
So you can sharpen 'just the eyes' if you slightly missed focus on them. Or selectively blur the front/back out of focus areas to get a stronger bokeh effect.
Since iOS 11 depth data is exposed in portrait photos allowing the focus point to be altered. It’s very effective.
You need an app though because it’s not possible in the Photos app (yet?) e.g. Focos
Apple bought PrimeSense - the company that owned the depth camera technology used in MS' Kinect - in 2013. I wouldn't be surprised if the iPhone uses a descendant of the Kinect.
IIRC, Lytro photos required special software (Flash/JS embed) in order to provide interactive exploration of the light field data. But looks like the specially hosted site, pictures.lytro.com, was shut down back in November 2017 [0], leaving users without a way to demonstrate the "Living Picture" functionality other than using the Lytro Desktop software, then exporting to GIF/MOV [1]. Anyone know if Lytro has mentioned plans to open source the format/software? I see they have a Github org with a bunch of repos, but most of it looks like forked libs and things related to devops: https://github.com/lytro
I thought that name sounded familiar. I really wanted one of those cameras when they came out, decided against when I saw all the software required to make it useful was closed. I had hoped they or someone else would release something but it never happened.
As far as I've been able to tell, their hosting service generated online viewers and embeds server-side, rather than their desktop software generating it and uploading it to pictures.lytro, Facebook, or 500px (the supporting services). It won't even let you do that kind of export without logging into api.lytro.com first. (Their Desktop 5 software is a .NET app if anyone wants to try poking at it.)
After their hosting service shut down last year, I crawled every embed the Internet Archive had, and downloaded the additional JSON and JPEG assets necessary to restore the embeds. https://www.archiveteam.org/index.php?title=Lytro
It's unclear whether all the Lytro file formats (including their later standard "XRAW") have any sort of documentation. The earlier formats (from their pre-Illum camera) do, e.g. https://github.com/nrpatel/lfptools but not the later formats, which include camera calibration data.
Oh wow! I had no idea anyone was actually making a digital lightfield camera, I thought sensor resolution was too small currently! It's a real shame they're shutting down but I guess there's no good way to view the images it captures.
Decades from now people will be writing blog posts about how lytro was ahead of it's time.
Sure, they made a good product and were ahead of their time, but software solutions for disparity maps are catching up. Light field arrays are very interesting science, and they definitely helped promote it, but as a business, well, they didn't treat their customers very well.
Anyone know whether this turned out to be an acquihire or not? Previous discussion [0] and articles [1] said Lytro was being acquihired by several of Google teams. But no exact price was mentioned ($25 to $40M).
Don't acquihire-closure announcements usually say something like, "We're proud to be joining the great team at Google, who shares our passion/commitment for this technology and [vague promise that this will somehow, at some point, benefit customers of the now-dead company]". But this announcement simply says:
> "We’re excited to see what new opportunities the future brings for the Lytro team as we go our separate ways"
I guess that's the least awkward way to phrase the situation if only some of the team got acquihired by Google?
But there's been very little information about how the acquisition fits into Google's strategy. I read somewhere that the Lytro staff that are brought across into Google will be dispersed across different departments. (obviously not everyone at Lytro was working directly on Lightfield technology but still...)
I am not into photography, but the idea of Lytro fascinated me! I distinctly remember that when it was launched, there was a layperson explanation of the concept along the lines of, "take a photo as you would normally do and, later on, decide which area to focus on, which to blur, etc."
As the cliché goes, "don't cry because it ended; smile because it happened."
I make lenticular artwork and owned a Lytro Illum before. The company was quite hostile after ending support for the camera, and refused to open source the web player, which was the only way you could make use of the living photos features and refocusing.
It is sad, because the Illum was a very well designed product and worked quite well for certain things. Unfortunately, Lytro was better at product design than business, and making the best feature of a $1500+ camera non-functional permanently was definitely a dick move.
I got an offer from Lytro in 2012 but I turned it down to stay at Google, where I still work.
At the time, I considered Lytro because it had real, substantial technology behind it, and was not just some trivial social networking app. I turned it down because assuming a $1B exit, my shares would still not have been that much more than what I got from Google. Six years later, I am doing much better (more earnings, more vacation, and probably better work/life balance) than if I had joined Lytro.
Although I made the right decision in hindsight, I think about how things might have been different if I had accepted that offer. It really calls into question the dominant Silicon Valley culture and narrative that we should join startups and change the world with new technologies.
> It really calls into question the dominant Silicon Valley culture and narrative that we should join startups and change the world with new technologies.
This sounds like a non sequitur. The main reason you joined Google over Lytro was money and quality of life.
There's nothing wrong with that. You didn't have as strong of a desire to change the world by building Lytro's new camera technologies. Others who were likely in similar situations balanced that equation differently.
"You didn't have as strong of a desire to change the world"
That's a rather dickish thing to say. Especially because, as we've seen time and time again, very, very few of these startups are actually "changing the world". More and more, being asked to join a startup is just being asked to overwork yourself for peanuts and then be screwed out of equity.
Ok, I understand what you mean now. I will still go on record as saying that, regardless of how strong you believe in something, you shouldn't take pay cuts to do it, especially given the record startups have for shafting employees. One really only has about 40 years to earn all the money they're going to earn, and that time can go by rather quick.
"I think about how things might have been different if I had accepted that offer. It really calls into question the dominant Silicon Valley culture and narrative that we should join startups and change the world with new technologies."
You would have made them succeed. Without you, they failed.
As someone who owned a Lytro Illumn, it was surprisingly a wonderful piece of hardware. Its massive and always gets interesting questions from onlookers. It was clear a lot of effort went into bringing a Shack–Hartmann wavefront sensor into the hands of a consumer. Sure it felt prototype-y, but it wasn't vaporware. It'd be wonderful if they opened up the development environment for the camera itself.
I really wanted to like my Lytro camera, and I was willing to put up with its hardware limitations, but I could never really trust it as a means of taking pictures I cared about so long as they were stuck inside Lytro's proprietary ecosystem.
58 comments
[ 4.8 ms ] story [ 128 ms ] thread"Instead of capturing it on a single plane, freezing an image in time and space, a light-field camera also captures the direction in which light was moving. Its processor then essentially renders a 3D scene, complete with the knowledge of distance between objects. A light-field photo represents not only everything in the scene, but a spatial understanding of the things in it." [1]
It's amazing from a technical perspective, but the image resolution was low (only 4 MP), and the price was high ($1600). [2]
I'd hoped that the technology would help the camera market find a new selling point to compete with phones, but apparently that hasn't happened. It's a pity that Lytro is disbanding instead of selling off to Canon or Nikon.
[1] https://www.theverge.com/2014/7/30/5949913/lytro-illum-revie...
[2] https://petapixel.com/2017/01/12/look-lytro-illum-camera-fut...
The Lytro embeds, unlike physical media (negatives/photos that get lost) could still be revived if Lytro would open source the code. The existing embeds might be dysfunctional, but I would hope the media data could be mirrored (from the Lytro servers) and played via embeds hosted on any server. I'm guessing Lytro photos never reached critical mass for people to really call for this archival effort.
Most webpages don’t have the reach of a mid-sized city newspaper in the 20th century. A few approach the reach of a major city newspaper, wire service like AP or Reuters, or radio/TV network – think about what percentage of the world could hear the BBC! Facebook or Google have those numbers but the vast majority of websites have far fewer visitors and less revenue per reader.
What’s different is that it’s much cheaper. You couldn’t run a printing faculty with a couple of people and laptops but a ton of websites do pull that off.
> The Lytro embeds, unlike physical media (negatives/photos that get lost) could still be revived
This leaves out a lot: if you have physical media, the risk is distributed and has been very well understood for centuries. If my house burns down, you don’t lose your family papers & photos. Important works are widely distributed and there’s rarely a single event which destroys all of them – even things like the Cultural Revolution didn’t manage 100% suppression.
Digital works are easier to make exact copies of but generally worse on every other aspect. In the case of a hosted service like Lytro, the complex proprietary stack is a huge risk but so is something as simple as archiving all of the data – if they’re doing any sort of on-demand loading you run the risk of only downloading the fraction of the data which displays initially without user interaction, and if they’re doing any sort of browser detection or rate-limiting you also have to be careful to deal with those. The web archiving community has years of experience with these kind of issues but it’s expensive and makes it hard for most people to archive their personal data.
If you look at the 1990 circulation numbers, they are small for mid-sized cities, and even for large cities. The NY Times circulation was just over a million, there are hundreds of webpages with that kind of reach now.
http://adage.com/images/random/0309/4-1990s-030909.pdf
They take a little while to load once you click on them; it's dozens and dozens of files.
The data for the embeds may still exist on Lytro's CDN, if you know the UUID for the image, and could be rescued and rehosted. After they shut their hosting service down last year, I crawled every embed the Internet Archive had, and downloaded the additional JSON and JPEG assets necessary to restore the embeds. https://www.archiveteam.org/index.php?title=Lytro
Edit: is there a reason why I can't download the torrent? Is it not available as a Torrent anymore?
Also note that I don't think the hosted versions can be imported back into the Lytro Desktop software, if that's what you were hoping for; they're just stacks of JPEGs, JSON metadata, and the depth map data.
Rehosting an individual embed should be pretty straightforward, just replacing the CDN URLs with relative or absolute ones to wherever it's been rehosted (or to Wayback Machine URLs). I'll be writing up some details at some point, but don't wait on me.
Rehosting an entire user's "pictures.lytro.com" gallery was trickier than that, I don't think I was able to get it working locally.
Good luck.
- the camera itself crashes about once every couple of minutes, and that is not an exaggeration. You literally need to keep putting the battery out and back in, and that is not isolated: every reviewer also mentions the same.
- the software generates a depth map from the light-field image. This depth map is often generated wrong, with blotches of foreground content being treated as the background, and vice versa. You have to manually export the depthmap as a grayscale image, fix it up in Photoshop, and import it back.
- when you share the image on lytro.com, it compresses this image and breaks it up into a few layers (presumably for the JS viewer), which significantly degrades the quality of the image
- the perspective effect isn't real, and uses content aware fill. This means that there are serious visual artefacts with this feature.
Ultimately Lytro will be a footnote in camera history. We’re already seeing scaled back implementations in SLRs and smart phones.
https://www.google.com/amp/s/photographylife.com/what-is-dua...
The new “Dual Pixel AF” sensors in some newer Canon DSLRs is much more like the Lytro, in that focus can be adjusted slightly in a single image capture, no burst capture trickery.
Reading a description of the Canon Dual-Pixel sensor feature https://www.canon.co.uk/cameras/eos-5d-mark-iv/dual-pixel-ra... ...it does not seem that it allows to adjust focus after the fact. It's something different again from Lytro and focus-bracketing.
As far as I can tell what it does is to capture depth information into the RAW file. I don't see how it could allow refocusing, but the examples show it allows you to selectively mask areas of the image based on distance from camera.
So you can sharpen 'just the eyes' if you slightly missed focus on them. Or selectively blur the front/back out of focus areas to get a stronger bokeh effect.
https://m.youtube.com/watch?v=HcQ9MSRRvn4
Also for a very large company, $40 million isn’t a lot.
https://itunes.apple.com/us/app/focos/id1274938524?mt=8
[0] https://support.lytro.com/hc/en-us/articles/115001259712
[1] https://support.lytro.com/hc/en-us/articles/115002733832
The Lytro founder's Ph.D thesis is extremely well written and informative on lightfield imaging as a whole.
Thank you, Lytro, for existing and showing us a new way forward. May the patents be licensed easily to keep the ball rolling!
As far as I've been able to tell, their hosting service generated online viewers and embeds server-side, rather than their desktop software generating it and uploading it to pictures.lytro, Facebook, or 500px (the supporting services). It won't even let you do that kind of export without logging into api.lytro.com first. (Their Desktop 5 software is a .NET app if anyone wants to try poking at it.)
After their hosting service shut down last year, I crawled every embed the Internet Archive had, and downloaded the additional JSON and JPEG assets necessary to restore the embeds. https://www.archiveteam.org/index.php?title=Lytro
Those were eventually added, so now any embeds the Wayback had before the shutdown should work again, e.g. http://www.theverge.com/2014/7/30/5949913/lytro-illum-review v. https://web.archive.org/web/20170317183900/http://www.thever...
It's unclear whether all the Lytro file formats (including their later standard "XRAW") have any sort of documentation. The earlier formats (from their pre-Illum camera) do, e.g. https://github.com/nrpatel/lfptools but not the later formats, which include camera calibration data.
The viewers on GitHub, e.g. https://github.com/damienfir/lytrogl and https://github.com/behnam/php-lytro are only for their earlier formats.
Decades from now people will be writing blog posts about how lytro was ahead of it's time.
Don't acquihire-closure announcements usually say something like, "We're proud to be joining the great team at Google, who shares our passion/commitment for this technology and [vague promise that this will somehow, at some point, benefit customers of the now-dead company]". But this announcement simply says:
> "We’re excited to see what new opportunities the future brings for the Lytro team as we go our separate ways"
I guess that's the least awkward way to phrase the situation if only some of the team got acquihired by Google?
[0] https://news.ycombinator.com/item?id=16705583
[1] https://www.theverge.com/2018/3/27/17166038/lytro-light-fiel...
I suspect a pure patent acquisition.
But there's been very little information about how the acquisition fits into Google's strategy. I read somewhere that the Lytro staff that are brought across into Google will be dispersed across different departments. (obviously not everyone at Lytro was working directly on Lightfield technology but still...)
The company wasn't making a lot of money and wasn't booming, otherwise it would not have been sold out, I reckon.
As the cliché goes, "don't cry because it ended; smile because it happened."
Maybe the photos are still viewable? There seems to be a bit of a controversy on that topic, but no one willing to say anything definite.
It is sad, because the Illum was a very well designed product and worked quite well for certain things. Unfortunately, Lytro was better at product design than business, and making the best feature of a $1500+ camera non-functional permanently was definitely a dick move.
Did the employees went to "greener pastures"?
At the time, I considered Lytro because it had real, substantial technology behind it, and was not just some trivial social networking app. I turned it down because assuming a $1B exit, my shares would still not have been that much more than what I got from Google. Six years later, I am doing much better (more earnings, more vacation, and probably better work/life balance) than if I had joined Lytro.
Although I made the right decision in hindsight, I think about how things might have been different if I had accepted that offer. It really calls into question the dominant Silicon Valley culture and narrative that we should join startups and change the world with new technologies.
This sounds like a non sequitur. The main reason you joined Google over Lytro was money and quality of life.
There's nothing wrong with that. You didn't have as strong of a desire to change the world by building Lytro's new camera technologies. Others who were likely in similar situations balanced that equation differently.
That's a rather dickish thing to say. Especially because, as we've seen time and time again, very, very few of these startups are actually "changing the world". More and more, being asked to join a startup is just being asked to overwork yourself for peanuts and then be screwed out of equity.
What I meant was the full sentence. GP specifically wasn't interested in what Lytro was building strongly enough to take the pay cut.
Agree the "we're changing/disrupting the world" meme gets misused.
I'll make an edit for clarity. Disrespect was not my intent.
Edit: the edit window on my original post just closed.
You would have made them succeed. Without you, they failed.
I want my pics.