RED Hydrogen review #hydrogen


gl
 


My brief review of the RED Hydrogen phone:

I just got try one.  I'll focus on the 3D aspects (I haven't used it as a phone):  I did just take it on a 12 mile walk though and put the on-board stereo camera through its paces:
 

- the screen is great overall.  Nice viewing angle based on 4 horizontal views (I never got on with untracked barrier displays and their horribly tight sweet-spots) - you still have to be in the right spot, but it's much more forgiving and natural, and the little look-around effect is nice when the content uses it well (see below).  Colours are good, and it's calm to look at (with good content).  And you don't see an obvious overlay pattern like with barrier displays.

Like other auto-stereo displays, it has limitations:

- reduced resolution to accommodate the views.  I don't find this critical when the content is awesome (worse when it isn't).

- ghosting (similar to other displays) when the deviation is too wide and image contrast is strong.

- significant brightness loss when the 3D 'backlight' engages (I assume there's a monochrome LCD overlay involved).

- apart from ghosting, maximum deviation is greatly limited by the way it works.  The 4 horizontal views cover the deviation - at small deviations you get pretty smooth view changes, giving the hologram-look-around like impression as you rotate the device.  But at larger deviations, the views are spaced further apart, and so the jumps are very noticeable - at some point it becomes distracting (as does the added ghosting) and ruins the illusion.

This means that you have to carefully manage the deviations of your content to get the best out of the display.  You can't get a lot of depth, but you can cheat a little - eg. blur the background, as that hides the obvious view transitions and so allows more deviation (sorry JR, shallow DOF is a plus on this screen : ).

Saying that, I've seen great content that understands how to get the best from the display.  That stuff really does look cool, and gives the hologram-in-your-hand impression (within a narrow but comfortable viewing angle).  There is magic in it.

But other content can be 'meh' even when there is some depth to it.

The big disappointment is the rear on-board stereo camera - the tiny lens separation produces little depth except for closer subjects.  Landscape shots often look very flat, and even closer subjects need careful composition to get a compelling 3D effect.

This is made worse by the software, which tries to 'optimise' the depth (and probably convergence) of shots after the fact, presumably to not exceed the display's deviation limits.  That seems smart - but in practice it is ultra-conservative, often producing so little depth that images that looked interesting on the display when you took them are flat as a pancake when you view them afterwards.

That's made worse by the generation of the in-between views - the software seems to rely on depthmaps not just to synthesize those, but also to 'optimise' the depth of the overall shot.  This means you see depth errors everywhere - it's really hard (impossible?) to get accurate depth maps from stereo content -  and images generally have a synthetic quality to them that, as a photographer, I don't like at all.  That might be reasonable if it had been aimed at the mass market, but it seemed to be pitched at content-creators like us, so a strange choice.

'Making 3D foolproof' is an admirable goal to get mass-market acceptance, but when the design choices make the resulting user-content uncompelling, that just makes everyone question (again) what the point of 3D actually is.

You _can_ edit your shots' depth and convergence (which the phone confusingly calls 'focus', a strange choice especially when lightfield cameras like Lytro already offer true post-shot focus adjustments) to make them more compelling.  I find this essential for pretty much every shot I've taken, and it works to a point - but it quickly looks synthetic, and you have to stay in reasonable deviation limits to avoid the jarring view transitions.  But it does give you scope to produce more interesting images.

The final nail in the coffin is that the camera's image quality is average at best - my ancient Samsung Note 4 phone takes better pictures.  There's also no zoom ability - my Note 4 only has quality-lowering software zooming, but that is crucial for framing shots with the right feel.

You also only get basic post-adjustments like brightness/contrast/saturation/hue.  Brightness and contrast are far too drastic, and there's no white-balance or sharpening  at all - again bizarre from a high-end camera maker! (hue adjustments are useless and don't replace white-balance, which is a very different operation).

In summary - when the content uses the display well, it looks really good.  The best I've seen is either rendered (ie. app-based CGI like games) or converted from other 3D sources.  Of course more resolution & wider viewing angles would be nice to have  (maybe we'll see those down the line).

You _can_ get interesting shots out of the on-board camera, but you have to compose the shot carefully and typically edit depth and convergence afterwards.

So, cool display, nicely designed handset, a front-facing stereo selfie cam that works well (as you tend to be close to it), but disappointing rear camera & processing that works against getting compelling depth from it.

But, very impressive that RED were able to make this phone at all - I can't imagine how many millions of things have to be worked out for that to happen for a first-time device, and how much it all cost them.  A second version would have been _really_ interesting, as I'm sure a lot of it would have improved.  As someone who takes tons of causal phone pics (as well as fine-art DSLR and video) I'd love a better quality 3D phone snapper with decent depth.
--
gl


 

My brief review of the RED Hydrogen phone:
Thanks for this review.

I just got try one.
Funny that -- I came to the same conclusion last week so I ordered one
on eBay. It arrives tomorrow. I plan on using it only as a 3D
camera, not a phone.

So, cool display, nicely designed handset, a front-facing stereo
selfie cam that works well (as you tend to be close to it), but
disappointing rear camera & processing that works against getting
compelling depth from it.
Have you tried using it in PRO mode to get just SBS images and thus (I
hope) avoiding the processing issues?

...BC


gl
 

On 09/07/2020 17:44, Bill Costa wrote:
My brief review of the RED Hydrogen phone:
Thanks for this review.
: )

Funny that -- I came to the same conclusion last week so I ordered one
on eBay.  It arrives tomorrow.  I plan on using it only as a 3D
camera, not a phone.
Yeah it's in the air (pun intended).

Have you tried using it in PRO mode to get just SBS images and thus (I
hope) avoiding the processing issues?
No I haven't tried taking the files off-cam yet.  I'm Windows-only here so I can't use the Mac software that allows view extraction. Are there other ways (I haven't researched that yet)?

I'm also wondering if H4V shots can be used as Facebook 3D pics somehow?  That would make them more interesting for me.
--
gl


turbguy
 

You can set the camera to capture "PRO SBS" mode in the camera  menu when in 3D capture mode.  Then download the files into your windows computer, then do what you need to do with the SBS file.  Mac not required.

Wayne

On Thu, Jul 9, 2020, 10:59 AM gl <gl@...> wrote:

On 09/07/2020 17:44, Bill Costa wrote:
>> My brief review of the RED Hydrogen phone:
> Thanks for this review.

: )

> Funny that -- I came to the same conclusion last week so I ordered one
> on eBay.  It arrives tomorrow.  I plan on using it only as a 3D
> camera, not a phone.

Yeah it's in the air (pun intended).

> Have you tried using it in PRO mode to get just SBS images and thus (I
> hope) avoiding the processing issues?

No I haven't tried taking the files off-cam yet.  I'm Windows-only here
so I can't use the Mac software that allows view extraction. Are there
other ways (I haven't researched that yet)?

I'm also wondering if H4V shots can be used as Facebook 3D pics
somehow?  That would make them more interesting for me.
--
gl




Nima
 

Great review!

it's really hard (impossible?) to get accurate depth maps from stereo content -  and images generally have a synthetic quality to them that, as a photographer, I don't like at all.  That might be reasonable if it had been aimed at the mass market, but it seemed to be pitched at content-creators like us, so a strange choice.

For content creators, there's the "quad" image format, which is a PNG or JPG in a 2x2 configuration.  If you create quad content, there will be zero post-processing done to it after the fact and will be displayed in LeiaPlayer and Holopix as-is.

I hope to see you share your content on Holopix in the future! 


JackDesBwa|3D
 

Thanks very much for your nice review!
I imagined it exactly like that with the technical elements that I knew, but it did not match with other descriptions which let me puzzled. Your observations totally make sense to me.

I'm Windows-only here
so I can't use the Mac software that allows view extraction. Are there
other ways (I haven't researched that yet)?

You can do it with exiftool that has a Windows version.
See https://photo-3d.groups.io/g/main/message/125625 for the arguments I used on Linux.

JackDesBwa


gl
 

On 09/07/2020 18:15, turbguy wrote:
You can set the camera to capture "PRO SBS" mode in the camera  menu when in 3D capture mode.  Then download the files into your windows computer, then do what you need to do with the SBS file.  Mac not required.
Yeah I've been shooting in Pro SBS, just not actually tried to see how you get the views from that.  Will give it a shot.
--
gl


gl
 

On 09/07/2020 22:13, Nima wrote:

Great review!
Thanks Nima.

For content creators, there's the "quad" image format, which is a PNG or JPG in a 2x2 configuration.  If you create quad content, there will be zero post-processing done to it after the fact and will be displayed in LeiaPlayer and Holopix as-is.
Do you mean that's if you author your own stereo content on another device?  Or can you shoot in quad on the phone?

If authored off-device, then with stereo content you'd need a way to synthesize the additional two views.
--
gl


gl
 


On 09/07/2020 23:46, Stereopix Net wrote:
Thanks very much for your nice review!
I imagined it exactly like that with the technical elements that I knew, but it did not match with other descriptions which let me puzzled. Your observations totally make sense to me.

Thanks, glad it came across OK.

You can do it with exiftool that has a Windows version.
See https://photo-3d.groups.io/g/main/message/125625 for the arguments I used on Linux.

Sweet, will have a play soon.
--
gl


Jay Kusnetz
 

If I am shooting on a slidebar, and want the normal stereo pair to be shot at 60mm apart, how should the 2x2 be shot? Do I shoot all 4 at 60mm apart, or would it be better to do 30-60-60-30 or 120-60-60-120?


Nima
 

Do you mean that's if you author your own stereo content on another device?  Or can you shoot in quad on the phone?If authored off-device, then with stereo content you'd need a way to synthesize the additional two views.
This is correct.  It hasn't seemed to be a problem for most of the RED community who's been making such content, because they either render it in Houdini or something similar, or take a stereo on Hydrogen or another rig and then composite it into four images in Photoshop.
If I am shooting on a slidebar, and want the normal stereo pair to be shot at 60mm apart, how should the 2x2 be shot? Do I shoot all 4 at 60mm apart, or would it be better to do 30-60-60-30 or 120-60-60-120?
Should each be equidistant, so 60mm apart each.


gl
 

On 10/07/2020 10:00, Nima wrote:
Do you mean that's if you author your own stereo content on another device?  Or can you shoot in quad on the phone?If authored off-device, then with stereo content you'd need a way to synthesize the additional two views.
This is correct.  It hasn't seemed to be a problem for most of the RED community who's been making such content, because they either render it in Houdini or something similar, or take a stereo on Hydrogen or another rig and then composite it into four images in Photoshop.

That doesn't explain how the extra two views are generated from the 2 view stereo content.  How are people doing that?  I'm sure there are options out there, but any recommended freeware with decent results?

> For content creators, there's the "quad" image format, which is a PNG or JPG in a 2x2 configuration.  If you create quad content, there will be zero post-processing done to it after the fact and will be displayed in LeiaPlayer and Holopix as-is.

Same question : ).

I've now played with getting my own Leia-screen optimised SBS content on using the "_2x1 " SBS image format, and it does mess with the depth and convergence (killing almost all depth for that specific image and moving the convergence).  It really would be better to have a 2x1 format that doesn't adjust anything, as then the phone software generates the in-betweens for you.

I think auto-generating them for photographic 2 view content makes much more sense than requiring every author (who cares to stay in control of their depth/convergence) to do it.

But after adjusting the depth & convergence manually, I'm getting a really nice result.
--
gl


gl
 

Nima, another question - if I edit depth and convergence (or actually any other image adjustments) inside the Leia app, save,, and then distribute the modified image to another H1 user, are those adjustments preserved, ie. will they see it as I did?

In technical terms, are the adjustments stored directly in the image file as metadata or in a separate phone-specific database?
--
gl


Nima
 

That doesn't explain how the extra two views are generated from the 2 view stereo content.  How are people doing that?  I'm sure there are options out there, but any recommended freeware with decent results?

If you're talking about how they can turn stereo photos into quad photos, I think it's probably just professional-level image composition in PhotoShop.  Once you identify what you want from the image, you can select the objects you want, overlay each view, see how far the natural disparity is in pixels between the same object in each view, and then copy the views and duplicate the object twice and move it the same amount for each exterior view(assuming you want your original stereo pair to be the inner views, do the opposite for the outer views).  In the past, we also allowed exporting from LeiaPlayer(back when it was RED Player) to quad, but then the extra views you were getting were synthesized by the image synthesis stack so that's probably not what purists are looking for.  I personally am not a PhotoShop expert, but this is what I've gathered from the Creators sharing their techniques on the forums.  For more details, feel free to ask on the H4VUser forums or the Leia Forums.



I've now played with getting my own Leia-screen optimised SBS content on using the "_2x1 " SBS image format, and it does mess with the depth and convergence (killing almost all depth for that specific image and moving the convergence).  It really would be better to have a 2x1 format that doesn't adjust anything, as then the phone software generates the in-betweens for you.

I think auto-generating them for photographic 2 view content makes much more sense than requiring every author (who cares to stay in control of their depth/convergence) to do it.But after adjusting the depth & convergence manually, I'm getting a really nice result.

The software makes a best guess assessment of the image, and obviously it may not work out great for images from unknown cameras or custom rigs with large baselines, as the software is optimized for images shot on Hydrogen.  Hopefully through editing you can get what you're looking for out of it.  

another question - if I edit depth and convergence (or actually any other image adjustments) inside the Leia app, save,, and then distribute the modified image to another H1 user, are those adjustments preserved, ie. will they see it as I did?

In technical terms, are the adjustments stored directly in the image file as metadata or in a separate phone-specific database?

In technical terms, Leia Image Format images are .jpg's that are encoded with metadata that does store all that information internally.  When shared to a standard 2D device, it will look like a regular 4K 2D image!  But when viewed on any Leia device, all the edits you've made will be preserved.  That said, be careful: many messaging applications strip image metadata to save themselves on bandwidth costs, so the metadata may not survive the transfer.  If you're texting the other person, Verizon Messages(works on any Android phone regardless of carrier) was designed specifically to support the transfer.  Full-sized image transfer through email, Telegram, and some others also works great.  Mixed results with Slack, Facebook Messenger, etc.


gl
 

Nima, having read my inline replies after writing, I figure I should explain where I'm coming from.  I'm investigating producing content from stereoscopic photography (both still and video) specifically authored for the H1 display.  That means I've already optimised the hell out of depth & convergence off-device, and don't want the software to change those.  But I do want the option to have the phone (or maybe some standalone software that Leia could provide) to generate the in-between views for me (see below for why):

On 16/07/2020 04:10, Nima wrote:

If you're talking about how they can turn stereo photos into quad photos, I think it's probably just professional-level image composition in PhotoShop.  Once you identify what you want from the image, you can select the objects you want, overlay each view, see how far the natural disparity is in pixels between the same object in each view, and then copy the views and duplicate the object twice and move it the same amount for each exterior view(assuming you want your original stereo pair to be the inner views, do the opposite for the outer views).
That is a lot of slow, error-prone work - you typically have the generate content that isn't visible in either view, when closer objects are obstructing the background.  This isn't trivial work unless you hack it.  And hacks may not look any better (or worse) than what your software already does.

High-quality manual work like that may be OK for occasional fine art shots, but the effort involved isn't really justified in general as it's only useful for this particular (discontinued) platform. Having the option of getting the views generated by your software without adjusting the original views would be much better to encourage high quality off-device generated content.

With my specially-authored "_2x1" test content, yes there are some depth errors, but (at least on that shot) they are not that important and everything looks great overall.  For worse errors, sure, you may want to hand edit, but I find that if the source content is compelling, minor artifacts aren't a big a deal.

 In the past, we also allowed exporting from LeiaPlayer(back when it was RED Player) to quad, but then the extra views you were getting were synthesized by the image synthesis stack so that's probably not what purists are looking for.  I personally am not a PhotoShop expert, but this is what I've gathered from the Creators sharing their techniques on the forums.  For more details, feel free to ask on the H4VUser forums or the Leia Forums.
Exporting the extra views makes sense -  you can always edit/replace them if you want.  Even if they contain problematic artifacts, it may be faster to edit those areas in Photoshop than creating whole new views from scratch.

So for all those reasons,  as you already generate them, why not make them available for SBS formats that also don't change depth & convergence (and also export them)?

The software makes a best guess assessment of the image, and obviously it may not work out great for images from unknown cameras or custom rigs with large baselines, as the software is optimized for images shot on Hydrogen.  Hopefully through editing you can get what you're looking for out of it.
In my case the stereo images were already ptimized for the Leia display, I'm authoring to get the absolute best result on it.  So I don't want the software to change anything (other than add the extra views).

In technical terms, Leia Image Format images are .jpg's that are encoded with metadata that does store all that information internally.  When shared to a standard 2D device, it will look like a regular 4K 2D image!  But when viewed on any Leia device, all the edits you've made will be preserved.
Perfect.

 That said, be careful: many messaging applications strip image metadata to save themselves on bandwidth costs, so the metadata may not survive the transfer.  If you're texting the other person, Verizon Messages(works on any Android phone regardless of carrier) was designed specifically to support the transfer.  Full-sized image transfer through email, Telegram, and some others also works great.  Mixed results with Slack, Facebook Messenger, etc.
Good to know, thanks.
--
gl


Nima
 

That is a lot of slow, error-prone work - you typically have the generate content that isn't visible in either view, when closer objects are obstructing the background.  This isn't trivial work unless you hack it.  And hacks may not look any better (or worse) than what your software already does.

This was my guess just based on what others have said on the forums.  Here are some topics sharing this process from Creators: 

https://h4vuser.net/t/converting-a-2d-image-to-h4v/8937

https://h4vuser.net/t/fcpx-works-nicely-for-editing-workflow-not-too-painful/10309

High-quality manual work like that may be OK for occasional fine art shots, but the effort involved isn't really justified in general

Like I said, other Creators are doing it, and it seems to be working fine for them.  Though I agree, the easier it can be for less technical Creators, the better.  There will be more simple tools coming out which will hopefully allow this in the future.  My plan is to bring the ability to generate high-quality quads from stereo into our Lightfield Studio software suite.

as it's only useful for this particular (discontinued) platform.

This is in no way a discontinued platform.  The Hydrogen hardware is discontinued, but everything on Hydrogen is forward-compatible with all LitByLeia devices.  If anything, Hydrogen was a Beta test.  The real first Leia product hasn't even come out yet!  

https://www.linkedin.com/posts/leia-inc-_lumepad-3d-lightfield-activity-6689406889966604288-hHy9

So for all those reasons,  as you already generate them, why not make them available for SBS formats that also don't change depth & convergence (and also export them)?

The main reason today is that our analytics show that SBS 3D fans are a tiny fraction of the total user base, and SBS 3D creators are even smaller.  However, I myself am a huge SBS 3D fan which is why I interact with the 3D communities, including this one.  I personally want more fans to use our products which makes it easier to justify additional resources put towards supporting them.  Your feedback has been heard though, and I think it aligns with what I've wanted us to do with Lightfield Studio, though it may take some months before we add this feature in.  In all transparency, it's probably likely sometime in the first half of 2021.

Really excited to see the content you're creating.  We'd love to have you sharing the content and asking more questions on the Leia Forum, where other members of the team can give you even better support than I'm able to, and your voice will be heard by many more people than just myself.

https://forums.leialoft.com


Olivier Cahen
 

This Red Hydrogen has a stereo base very short, so that it can shoot only stereo pictures of near objects. I would never buy such a junk.

Le 16 juil. 2020 à 23:31, Nima <contactnimaz@...> a écrit :
That is a lot of slow, error-prone work - you typically have the generate content that isn't visible in either view, when closer objects are obstructing the background.  This isn't trivial work unless you hack it.  And hacks may not look any better (or worse) than what your software already does.

This was my guess just based on what others have said on the forums.  Here are some topics sharing this process from Creators: 

https://h4vuser.net/t/converting-a-2d-image-to-h4v/8937

https://h4vuser.net/t/fcpx-works-nicely-for-editing-workflow-not-too-painful/10309

High-quality manual work like that may be OK for occasional fine art shots, but the effort involved isn't really justified in general

Like I said, other Creators are doing it, and it seems to be working fine for them.  Though I agree, the easier it can be for less technical Creators, the better.  There will be more simple tools coming out which will hopefully allow this in the future.  My plan is to bring the ability to generate high-quality quads from stereo into our Lightfield Studio software suite.

as it's only useful for this particular (discontinued) platform.

This is in no way a discontinued platform.  The Hydrogen hardware is discontinued, but everything on Hydrogen is forward-compatible with all LitByLeia devices.  If anything, Hydrogen was a Beta test.  The real first Leia product hasn't even come out yet!  

https://www.linkedin.com/posts/leia-inc-_lumepad-3d-lightfield-activity-6689406889966604288-hHy9

So for all those reasons,  as you already generate them, why not make them available for SBS formats that also don't change depth & convergence (and also export them)?

The main reason today is that our analytics show that SBS 3D fans are a tiny fraction of the total user base, and SBS 3D creators are even smaller.  However, I myself am a huge SBS 3D fan which is why I interact with the 3D communities, including this one.  I personally want more fans to use our products which makes it easier to justify additional resources put towards supporting them.  Your feedback has been heard though, and I think it aligns with what I've wanted us to do with Lightfield Studio, though it may take some months before we add this feature in.  In all transparency, it's probably likely sometime in the first half of 2021.

Really excited to see the content you're creating.  We'd love to have you sharing the content and asking more questions on the Leia Forum, where other members of the team can give you even better support than I'm able to, and your voice will be heard by many more people than just myself.

https://forums.leialoft.com



turbguy
 


"This Red Hydrogen has a stereo base very short, so that it can shoot only stereo pictures of near objects. I would never buy such a junk".

True, it could be larger,but it does have it's uses.  And you get a whole lot more for your money than just a true stereo camera with a good 3D screen.

Wayne



robert mcafee
 

We need someone to design and fabricate a stereo base expander an Cyclopital had done for the Fuji W3!

On Friday, July 17, 2020, 10:14:10 AM EDT, turbguy <wkarberg@...> wrote:



"This Red Hydrogen has a stereo base very short, so that it can shoot only stereo pictures of near objects. I would never buy such a junk".

True, it could be larger,but it does have it's uses.  And you get a whole lot more for your money than just a true stereo camera with a good 3D screen.

Wayne



Nima
 

This Red Hydrogen has a stereo base very short, so that it can shoot only stereo pictures of near objects. I would never buy such a junk.

If you're looking exclusively for a stereo SBS camera to create images for traditional 3D displays, I agree it's not super useful except for close-up shots.  

However, Hydrogen is it's own 3D platform.  It has a comfortable display, SDK, apps, games, and community.  It computationally estimates depth to generate multiscopic, not stereoscopic images.  e.g. the low stereo baseline is irrelevant for the consumer use cases the product is designed for.  If stereoscopic 3D photographers don't like it...oh well 💁🏽 they're not the target demographic.  I want us to support 3D enthusiasts but if the resounding statement from the community is that they don't like it and don't care, then that's fine too.  I'm trying my best to take care of the 3D community, and that's all I could hope for.  Currently we support backwards compatibility of SBS images and the ability to shoot in SBS on our products, but if the community truly doesn't want it/doesn't care then those features may be deprecated in the future to provide for a cleaner user experience for our core customers.

We need someone to design and fabricate a stereo base expander an Cyclopital had done for the Fuji W3!

I would LOVE if someone did that.  If they do, just tell me where to send my money!  I'd pay through the nose for that.  I'd love something like the BeastGrip with a stereo base extender lens mount.  Maybe even something that mounts to BeastGrip?