Here we go, finally our first 4K RAW video test made with the Canon 5D Mark III. Following our 5D Mark II previous tests, we have exported the dng sequence in 3840x 2160 jpeg files directly from Lightroom. No moire and aliasing. We will make a tutorial about all of this soon, stay tuned!
Second 4K RAW video attempt made with the Canon 5D Mark III. Following the previous handheld test http://youtu.be/gONtARNBrbw we have now shot on tripod, in order to get as much details as possible. The main experiment is to resize the 1920x1080 frames to 3840x 2160 from Lightroom.
IMPORTANT: Having analyzed and compared the original files with the re-sized ones, we have found no difference at all in terms of details. In conclusion, up-scaling the files is not going to actually "resuscitate" any "lost detail", not at all. But for sure, when we upload the video on Youtube (or Vimeo) and we then compare it with the HD version of it, it is clear enough that the "4K" version offers a superior and impressive quality, even when it is played in HD resolution. So, nothing revolutionary as some people is suggesting, unfortunately not at all. But worth a try. This is our definitive thought.
All the best.
The video has been shot on:
Canon 5D Mark III
Tamron SP 70-200mm f2.8 USD
CF Lexar Professional 1000x 32GB
Special thanks to EXIV
Music: "Yiourgh" by DoKashiteru (feat. Coblat)
http://ccmixter.org/files/DoKashiteru/21394
is licensed under a Creative Commons license:
http://creativecommons.org/licenses/sampling+/1.0/
The video has been shot on:
Canon 5D Mark III
Tamron SP 70-200mm f2.8 USD
CF Lexar Professional 1000x 32GB
Special thanks to EXIV
Music: "Yiourgh" by DoKashiteru (feat. Coblat)
http://ccmixter.org/files/DoKashiteru/21394
is licensed under a Creative Commons license:
http://creativecommons.org/licenses/sampling+/1.0
So far so good :D
My monitor is 2k max :D
Quotewe have exported the dng sequence in 3840x 2160 jpeg files
I can export my dng sequence in 8k ;)
I should have read the description. The topic title is misleading. Yeah actually with upscaling you can achieve unlimited Resolution, haha.
Quote from: pascal on October 28, 2013, 12:05:55 AM
I should have read the description. The topic title is misleading. Yeah actually with upscaling you can achieve unlimited Resolution, haha.
But when you have the video on Youtube or Vimeo (that is the place where a corporate or a video like this is going to be) the "4K" video is actually and drastically better than the 1920x1080 version. So it make sense to test it and see if anything good can come from it. No miracles promised, just trying to push the footage at a relevant level. All the best.
2014-03-16 Update: the up-scaling experiment may really help when you plan to upload a video on the internet, since the HD version you get by uploading a 4K version of it (for example on Youtube) is always better then the one you get by uploading a video processed in HD (as all of us usually do). Here is a normal HD raw video, which I have just shot with the new ML version for the 5d Mark III 1.2.3 you can really see this is not as good as the HD version of the 4K tests. Because of the way Youtube compress the video, this seems less sharp than the one of the video posted above.
What's the point of this video? This is not a 4k raw attempt... Just footage scaled to 4k....I guess YouTube views??
Quote from: pascal on October 28, 2013, 12:05:55 AM
I should have read the description. The topic title is misleading. Yeah actually with upscaling you can achieve unlimited Resolution, haha.
You're totally right, Pascal !
But I like the idea of testing this aspect...
Because I've tried to compare 30 secondes of video shoot with a Canon 5D Mark3 in RAW .dng (1920x1080) and the same scene shoot with an Alexa 2K.
Then I've made a DCP and organised a comparaison in a small movie theater....
To be able to make the DCP we have upscaled the Magic Lantern Footage to 2k, and I was curious to see the result.
I've made some drastic modifications on the exposition and drastic color correction to compare the capacity of the 2 différent files (ARRIRAW and ML RAW)...
The result is just amazing.
The upscale does not destroy the quality of the ML file, it gives just a bit of sotness, wich is good for my point of view compare to the ARRIRAW wich is naturally soft.
Resolution, lattitude are same !
More than that, the upscale gives a cine look to the image, compare to the "surgical precision" of the RED...
Thank you EXIV, your idea is not bad.
Fran6
Quote from: Francis Frenkel on October 29, 2013, 10:32:49 AM
You're totally right, Pascal !
But I like the idea of testing this aspect...
Because I've tried to compare 30 secondes of video shoot with a Canon 5D Mark3 in RAW .dng (1920x1080) and the same scene shoot with an Alexa 2K.
Then I've made a DCP and organised a comparaison in a small movie theater....
To be able to make the DCP we have upscaled the Magic Lantern Footage to 2k, and I was curious to see the result.
I've made some drastic modifications on the exposition and drastic color correction to compare the capacity of the 2 différent files (ARRIRAW and ML RAW)...
The result is just amazing.
The upscale does not destroy the quality of the ML file, it gives just a bit of sotness, wich is good for my point of view compare to the ARRIRAW wich is naturally soft.
Resolution, lattitude are same !
More than that, the upscale gives a cine look to the image, compare to the "surgical precision" of the RED...
Thank you EXIV, your idea is not bad.
Fran6
Thanks a lot Francis, I really support what you just said, also considering that when you open the file sequence on Lightroom you can also increase the sharpness of at least 60% (which is still amazing) without damaging the "softness" of the image. You see how testing with this technology is so important for all of us. By the way, please have a look at a more "tripod" kind of test made on the same landscape, and read my definitive thoughts about all of this.
Second 4K RAW video attempt made with the Canon 5D Mark III. Following the previous handheld test http://youtu.be/gONtARNBrbw we have now shot on tripod, in order to get as much details as possible. The main experiment consist in resizing the 1920x1080 dng frames to 3840x 2160 from Lightroom.
IMPORTANT: Having analyzed and compared the original files with the re-sized ones, we have found no difference at all in terms of details. In conclusion, up-scaling the files is not going to actually "resuscitate" any "lost detail", not at all. But for sure, when we upload the video on Youtube (or Vimeo) and we then compare it with the HD version of it, it is clear enough that the "4K" version offer a superior and impressive quality, even when it is played in HD resolution. So, nothing revolutionary as some people is suggesting, unfortunately not at all. This is our definitive thought.
Quoteyou can also increase the sharpness of at least 60% (which is still amazing) without damaging the "softness" of the image
I dont think so EXIV...
On your exemple the sharpening is too much visible on the upscale sequence.
> Try to make the same experiment on a face, with hair on a solid color background and it will be visible and it will look ugly.
If you sharpen to much on the original image, and then upscale after : you upscale the "sharpening lines" on the borders of the objects in the image...and make it visible.
Try to sharpen a bit after (if you need too sharpen), not before.
Francis
Quote from: Francis Frenkel on October 31, 2013, 08:01:44 AM
I dont think so EXIV...
On your exemple the sharpening is too much visible on the upscale sequence.
> Try to make the same experiment on a face, with hair on a solid color background and it will be visible and it will look ugly.
If you sharpen to much on the original image, and then upscale after : you upscale the "sharpening lines" on the borders of the objects in the image...and make it visible.
Try to sharpen a bit after (if you need too sharpen), not before.
Francis
Hi Francis, thanks a lot for the suggestion. With the Canon 5D Mark II I couldn't really try to attempt that much, but with the Mark III it is actually working. By the way, have you got any idea of when will be available an official ML version for the Mark III? I really would like the possibility to shoot more than HD in RAW...
Guys, why don't you make some test as well and then post it here, so we all can watch?! :)
Quote from: EXIV on October 31, 2013, 01:19:31 PM
By the way, have you got any idea of when will be available an official ML version for the Mark III?
Like Canon taking ML and releasing it as firmware?
Quote from: EXIV on October 31, 2013, 01:19:31 PM
By the way, have you got any idea of when will be available an official ML version for the Mark III?
http://wiki.magiclantern.fm/faq
The very bottom of the page has the release date ;D
Stunning!
Wow. Awesome stuffs
I get great results exporting to 3k+ an than downscaling to 1080 on my MKII. I'm actually quite frustrated that ACR 8.2 does not give you "suggestions" of up and downscale anymore. Before it would give you from 3 to 5 different default options. On an usual MKII cr2 file would be 21mp (default) 25mp (upscale) and 18mp and 12mp downscale. Now you have to do the math yourself. What the f... Is wrong with this adobe people anyway?
This is cool because 600D and other SD cams have to make do with less and always upscale. I guess this is what happens when you do it to 1080P footage, wonder how 50D and 7D would compare.
QuoteNow you have to do the math yourself. What the f... Is wrong with this adobe people anyway?
I know, I enjoyed default up-scaling options too.
Does ACR ver 7 have upscale feature? I cannot find it anywhere...
Quote from: kgv5 on November 05, 2013, 09:26:10 AM
Does ACR ver 7 have upscale feature? I cannot find it anywhere...
(http://ipic.su/img/img7/fs/kiss_34kb.1383641611.jpg)
Do you guys not see this or something? ???
(http://i.imgbox.com/adcrBuQp.jpg)
Quote from: Andy600 on November 05, 2013, 12:44:56 PM
Do you guys not see this or something? ???
(http://i.imgbox.com/adcrBuQp.jpg)
Andy, would you like to make a test as well and share it here? We would love to see others experimenting on this.
There is no point TBH. If it ain't there in the original shot you can't conjure it up through scaling. I've tried scaling raw HD to 4K and beyond with ACR, resolve, AE, instant HD etc and I can't see any 'real' increase in resolution, sharpness or detail. The only 'perceived' increase in detail you can achieve is if you use tone mapping. Also, increasing sharpness beyond a few % is not the way to go for natural looking shots IMO. If you want 4k resolution you have to shoot native 4K.
The only possible benefit in upscaling is for Youtube playback of footage shot at or upscaled to 2K/4K at 1080p or if you have to upscale for a 2K+ theater viewing.
sorry for being such a lay-man here, and sorry if somebody already posted an answer to this and I missed it but:
is this really real 4k footage with ML on a 5d?
how large are the original raw-frames anyway? I read somewhere that they're 4368x2912, which is just about 4k (actually larger right?), so in that case - why is there even a thread about this like it's something new, since shooting RAW is basically the main purpose for ML at all?
I dont have ML on my camera yet, but I'm thinking about trying it out, why I'm snooping around here, and asking stupid questions...
second:
if this is not recorded 4k footage, and just regular 2k (?) upscaled (i.e. blown up) to cover a 4k "frame area", then by all definitions it's not 4k at all and this is just silly, and it would definately not gain any quality... so what am I missing? what are we talking about here?
Those raw frames you are thinking about are the images captured in photo. ML raw saves a smaller part from live view (~1920). My actual frames are 1920 x 818.
What he means by 4k is ~1920 scaled in ACR to 4096. It handles upres fairly well, and technically it is 4k resolution.
I like to boost to 2k in ACR before processing. I don't see a noticeable dip in quality, and it gives me more info when I get to export. Youtube absolutely destroys videos when it compresses. Beefy 2k files look more like true 1080p on youtube imo ;)
edit: it is a bit silly. I suppose if you shot a pickup on ML-raw and needed to place it into an existing 4k timeline (with some red footage?) that it might be a good idea. Or if you needed 2k master.
thanks for the answer! but...
you need to explain this as if to a three-year old.
I'm serious. And I need to get this.
Acronyms? I'm lost.
So, if you please... I'd really appreciate it!
Thank you!
Simply put, the 5d does not shoot 4 K. the original post was an "experiment" to see if they can stretch his footage to 4k.
he said; "The main experiment consist in resizing the 1920x1080 dng frames to 3840x 2160 from Lightroom. "
Its like back in the day when you would stretch SD footage to match them with 1080 footage on a 1080 timeline.
Quote from: EXIV on October 27, 2013, 07:19:35 PM
Here we go, finally our first 4K RAW video test made with the Canon 5D Mark III. Following our 5D Mark II previous tests, we have exported the dng sequence in 3840x 2160 jpeg files directly from Lightroom. No moire and aliasing. We will make a tutorial about all of this soon, stay tuned!
Actually this was quite annoying. And really quite confusing for most people. ???
IMHO this has nothing to do with ML 4k RAW footage.
You should consider renaming the title of this thread as it is now receiving a lot of pointless attention.
And, do you really think you are the first person to upscale 2k footage? Like the FIRST EVER attempt, really? Pretty naive.
How about this topic: "Another attempt at upscaling 2K RAW Canon 5D Mark III footage in ACR"
Thanks.
You can record 4K with 5D3 if you lower frame rate. Please see test here:
http://www.magiclantern.fm/forum/index.php?topic=12253.0 (http://www.magiclantern.fm/forum/index.php?topic=12253.0)
Lenses:
Kowa 8z lens
Canon 85mm lens
Settings:
Anamorphic x2 settings: 1536x1290 24p
3K cropped: 2880x1206 12p
4K cropped: 3584x1320 12p
The Komputerbay 64GB card does not go beyond 112 MB/s (debug mode), this is why in 4K and 3K framarate has to be lowered.
Enjoy!
Modify message
« Last Edit: June 08, 2014, 08:01:10 PM by jgerstel »
And here another 3K vid with 5D3 - korfball match with my kids
Anamorphic x2 settings: 1536x1290 24p
What's the purpose of shooting with anamorphic lens? I did a bit of research to understand what is it and why it was developed (to better cover the sensor area), but in this hack the limitation is caused by the resolution, so we cannot use the full sensor area anyway.
I also found that often anamorphic lenses are just normal lenses with an anamorphic element adding potential distortion and reducing light transmission.
Thanks