Luminous Landscape Forum
Equipment & Techniques => Motion & Video => Topic started by: MichaelEzra on May 25, 2019, 10:21:23 pm
-
I decided to stick with Davinci Resolve and am looking into a higher bit rate recording for video. Since I am using Nikon Z7, the only current option is Atomos recorder with N-Log and Prores Raw in the future. The latter is not an option for me as I am on Windows and it is unlikely that it will ever be licensed for Windows. This made me look into BMPCC 4K that just got upgraded with the B-Raw capability with its well optimized workflow with Resolve.
While BMPCC 4K is a cinema camera, Nikon Z7 has a very convenient auto focus and high ISO capability.
Aside from the understandable camera system differences, I am curious how the image quality and workflow with B-Raw/Davinci Resolve compare to N-log/Davinci Resolve.
-
Aside from the understandable camera system differences, I am curious how the image quality and workflow with B-Raw/Davinci Resolve compare to N-log/Davinci Resolve.
I can't comment on the image quality of these two cameras, since I don't shoot with either, but do know that the workflow is noticebly faster with B-Raw rather than 4:2:2 10-bit files due to the GPU acceleration option for the former (Studio version). 4:2:2 decoding is all on CPU.
It is 60fps for B-Raw vs 40fps for DNxHR 4:2:2 10-bit 4K DCI timeline playback, for example.
-
Thanks, this is great to know!
-
I decided to stick with Davinci Resolve and am looking into a higher bit rate recording for video. Since I am using Nikon Z7, the only current option is Atomos recorder with N-Log and Prores Raw in the future. The latter is not an option for me as I am on Windows and it is unlikely that it will ever be licensed for Windows. This made me look into BMPCC 4K that just got upgraded with the B-Raw capability with its well optimized workflow with Resolve.
While BMPCC 4K is a cinema camera, Nikon Z7 has a very convenient auto focus and high ISO capability.
Aside from the understandable camera system differences, I am curious how the image quality and workflow with B-Raw/Davinci Resolve compare to N-log/Davinci Resolve.
Michael, I also haven't used either of these cameras. But I have shot with a Sony A7rIII to ProRes on an external recorder. I did test the camera's internal recording for video and it was not very good. But, even with this method, it's difficult to grade the images and they fall apart rather quickly. And shooting video with a still camera drives me nuts as well. And I've learned never to shoot video on these cameras with the anti shake on!
Recording Log to an external recorder does give you full dynamic range, but, with the 8bit output, you will see some banding after color correction, especially in smooth skies or walls. So, if the Nikon only has 8bit output over HDMI, you will also have this issue.
As for the BM camera, from what I've seen on line, I'd view that as a much better "low cost" option for shooting motion. And there's now a 6K version that you might want to look into.
What ever camera you choose, it's way easier to shoot and focus motion with a lens designed for manual focus, preferably a cine style lens. Focusing a "fly by wire" auto focus lens for motion is an experience I'd rather not repeat. Either using auto focus or trying to manually focus it, which is very very difficult.
Let us know what you've chosen and how it worked out Michael!
Good luck :)
-
I can't comment on the image quality of these two cameras, since I don't shoot with either, but do know that the workflow is noticebly faster with B-Raw rather than 4:2:2 10-bit files due to the GPU acceleration option for the former (Studio version). 4:2:2 decoding is all on CPU.
It is 60fps for B-Raw vs 40fps for DNxHR 4:2:2 10-bit 4K DCI timeline playback, for example.
I have both the Z7 and the Z6, and have used both for video in exactly the way you mention. I also own the Sony A7r3 and have used it for video. I have no experience with the BMC cameras. My primary business is video, and my primary video camera is a Red Dragon.
First, the video output over HDMI from the Zs is 4:2:2 10 bit (assuming you set the camera th output 10 bit.) This is available for N-log or any of the other profiles. Recorded to ProRes 422HQ or the equivalent DnX variant is quite a robust file for grading. As with any non-raw file, it is important to get white balance right in camera. Shooting charts helps a lot.
The A7r3 sends only 8-bit out its HDMI port. It helps that it is 4:2:2, rather than the 4:2:0 it records internally, but the Nikon file is a LOT more gradable.
I find Resolve’s response just fine with the ProRes files, but I grade a lot of 6k Red raw, which is a very demanding codec for any system, so my perspective on that may be distorted. Note that there is a lot more to system performance in video that unpacking the codec. Data rates have a huge impact on an edit systems’ response, even more so if you start layering clips, so disk speed is massively important, as is RAM.
As for focus, if you want to go manual focus, the Nikon with an adapter can mount any lens you’d like. If you want to use PL-mount glass, the Z7 is probably a better choice than the Z6, because its apsc (super 35) output in 4K is better. The 6K BMC has an EF mount, the 4K version MFT. I don’t know enough about them to know what adapters are available.
If you want to use autofocus, the Zs with native lenses are really very good. The PSA linked below was shot on a Z6, with the 35mm and 50mm S lenses at f/2.0 and 2.2 in quite low light—just two practical incandescents. I think we shot at iso 1600. It was N-log to a Shogun. The camera was mounted on a gimbal, so it and the talent were constantly in motion. It would have taken a very skilled focus puller to match the performance of the autofocus in this scene.
https://vimeo.com/347771601
-
Bruce, David, thanks for the info. I guess what I'd like ultimately is Z7 with its existing video autofocus, but recording the open format blackmagick raw internally to the fast XQD cards.
And if heat generation is an issue, even APS-C crop limitation would be great. Wouldn't that be just amazing... oh well:)
I also noticed Z7 autofocus in video is extremely helpful. For a one man production, this becomes indispensable.. Its great to know that HDMI 10 bit is supported for all profiles in the Nikon - I found AlvaroYus-Curve with D-lighting gives a workable image, but is posterized in 8 bit recording (open blue sky), so hopefully 10-bit external recording resolves that. Using N-Log comes with a number of limitations (including usability of the autofocus), while using a custom picture profile does not. Z-log is also a possibility for this scenario.
David, have you done any comparisons of Z-log vs N-log on Z7?
-
Bruce, David, thanks for the info. I guess what I'd like ultimately is Z7 with its existing video autofocus, but recording the open format blackmagick raw internally to the fast XQD cards.
And if heat generation is an issue, even APS-C crop limitation would be great. Wouldn't that be just amazing... oh well:)
...
David, have you done any comparisons of Z-log vs N-log on Z7?
I only recently learned about Z-log, so I haven’t bought or tested it yet. (I’m currently on vacation sitting by the ocean, so it’ll be a week or more before I do that.)
I also find that turning view assist on helps autofocus when shooting log, but at the expense of a few more microseconds of delay in the finder image.
The biggest issue I have with N-Log is that Nikon has not, to my knowledge, published either a normalization LUT or a spec for it. I can grade it in Resolve quite well, but for any technical or VFX work, that isn’t really enough. Red’s REDLOG FILM, for example, is based on, and hews very closely to Cineon Log, as does Sony’s S-Log3. Those are easy to use when working with others. Arri Log is it’s own well-published standard. N-Log, like S-log2 is a “who knows” curve, which has real issues for professional work. Better than nothing, but really, not that much better. 10-bit, OTOH, is a Godsend.
As for the APS-C/Super-35 “limitation,” I’m fine with that. Having grown up using film, where S35 was large format, I consider it the norm, not a compromise at all. And the 14-30 with that crop is a truly brilliant lens. However, the Z6 is significantly better at focus in low light, if that matters to you, and I believe it is better at autofocus overall. I attribute that to the lower data rate—it’s no coincidence that all of the super-af cameras are in the 24 Mpix range: D5, 1D, A9, SL. I think the data processing requirements of the Z7 slow down the autofocus.
Is BM Raw open? That’s news to me. I’m very interested in what happens with ProRes Raw when Atomos and Nikon release that upgrade. Apple and BM haven't seemed to have had the best of relationships. And at the moment, I understand that Final Cut is the only software able to read ProRes Raw. I may be forced to buy the program just for that capability. (I’m using Macs, but not FCP.) It wouldn’t be out of the question for you to pick up a Mac Mini or IMac and FCP just to transcode ProRes Raw. If the camera works well for you, that’s a much more practical solution than it might seem at first blush.
-
On BM raw: Open Standard and Free to Download, Cross platform and license free! (https://www.blackmagicdesign.com/products/blackmagicraw)
It would really be ideal, I wonder if there is a way to petition Nikon for this?
Red and Apple are in dispute over the patent rights on Prores Raw... so who know when it happens with Nikon. But getting a whole Mac just to transcode pains me; this is the ugly wasteful truth of our time...
-
On BM raw: Open Standard and Free to Download, Cross platform and license free! (https://www.blackmagicdesign.com/products/blackmagicraw)
It would really be ideal, I wonder if there is a way to petition Nikon for this?
Red and Apple are in dispute over the patent rights on Prores Raw... so who know when it happens with Nikon. But getting a whole Mac just to transcode pains me; this is the ugly wasteful truth of our time...
Michael, "Cross platform and license free" applies to Braw decoder. Encoder SDK, however, is not available to developers outside BMD for free. Not sure, if they even offer it for a fee.
Frankly, I don't understand the hype over 10bit 4:2:2 encoded in ProRes RAW when the same quality signal is already available via ProRes and DNxHR codecs for those shooting on Nikon Z and Atomos Recorders.
-
Hi Mailna, on the same page it says "Free Developer SDK Fully documented SDK is publicly available!"
Raw should be 12 bit, not 10;) One of the advantages of ProRes/BM Raw is that it takes much smaller space in storage compared to ProRes/DNxHR.
-
Hi Mailna, on the same page it says "Free Developer SDK Fully documented SDK is publicly available!"
"Using the SDK gives you access to GPU accelerated and CPU optimized algorithms for decoding Blackmagic RAW files." I repeat, it is only for decoding .braw. There is nothing mentioned on that page about encoding tools. BMD is not giving away Braw encoding source code.
Raw should be 12 bit, not 10;) One of the advantages of ProRes/BM Raw is that it takes much smaller space in storage compared to ProRes/DNxHR.
Both Nikon Z & Ninja V mention 10-bit 4:2:2 output and input respectively over HDMI 2.0. Atomos recorder may spit out a 12-bit capable ProRes RAW. And one may choose to believe it is a 12-bit color depth source file, though both parties do not promise that. Does it really matter, especially for Windows users? :)
"Apple ProRes RAW data rates generally fall between those of Apple ProRes 422 and Apple ProRes 422 HQ." (Source: Apple_ProRes_RAW_White_Paper.pdf)
-
Both Nikon Z & Ninja V mention 10-bit 4:2:2 output and input respectively over HDMI 2.0. Atomos recorder may spit out a 12-bit capable ProRes RAW. And one may choose to believe it is a 12-bit color depth source file, though both parties do not promise that. Does it really matter, especially for Windows users? :)
"Apple ProRes RAW data rates generally fall between those of Apple ProRes 422 and Apple ProRes 422 HQ." (Source: Apple_ProRes_RAW_White_Paper.pdf)
At this point, nobody who’s talking knows what the Atomos/Nikon project will produce, but it matters a great deal. The two things that make raw files worthwhile are the ability to adjust white balance in post and the availability of greater color depth for grading than the various flavors of RGB that camera’s produce. The lowest bit depth the Nikon’s produce in raw still images is 12-bit. It is my hope that the ProRes raw file is also a 12 file. Oversampling matters a lot when grading, and anyone doing serious grading (regardless of whether on Windows, Mac, or Linux) is grading to a 10-bit, calibrated monitor.
So yes, it matters. If the Atomos/Nikon project produces 10-bit raw, that will be nice. But if they produce 12-bit raw, they will be producing class-leading files.
Comparing data rates between raw files and RGB files isn’t very enlightening without knowing what compression schemes, if any, are being used in the raw files. It’s normal for raw files to be considerably smaller than RGB files, other things being equal, but compression and bit depth affect both.
-
Oversampling matters a lot when grading, and anyone doing serious grading (regardless of whether on Windows, Mac, or Linux) is grading to a 10-bit, calibrated monitor.
10, 12, or even 14 bits can be important when capturing the original image and when performing color grading as you've said.
But, once the image has been graded, for the most part, an 8 bit output to one's display is very usable. (HDR video release excluded here). What is important is that the display be accurately calibrated. And for video that means not using the OS .icc profiles as I don't know of any video grading app that is color managed to use .icc profiles as Photoshop does.
I will not turn this thread into a "how to calibrate" for video thread, but I just wanted to point out that many many professional colorists do work with 8 bit video output to their displays, at least for SDR/HDTV/cinema work and it's fine. Viewing in 8 bit does not change the grade, but maybe at the margins, helps one see any banding artifacts, or lack of them.
-
10, 12, or even 14 bits can be important when capturing the original image and when performing color grading as you've said.
But, once the image has been graded, for the most part, an 8 bit output to one's display is very usable. (HDR video release excluded here). What is important is that the display be accurately calibrated. And for video that means not using the OS .icc profiles as I don't know of any video grading app that is color managed to use .icc profiles as Photoshop does.
I will not turn this thread into a "how to calibrate" for video thread, but I just wanted to point out that many many professional colorists do work with 8 bit video output to their displays, at least for SDR/HDTV/cinema work and it's fine. Viewing in 8 bit does not change the grade, but maybe at the margins, helps one see any banding artifacts, or lack of them.
I’m not sure what you’re saying here. Sure, 8-bit output to a monitor is useful after an image is graded, but for grading it’s a real compromise.
Maybe I work in different circles from yours, but I don’t know any professional colorists who work with 8-bit monitors for grading. Sure there are editors who do a bit of grading, but colorists? Not that I’ve seen. Viewing in 8-bit certainly does change a grade. To cite two simple ways it does, some dark nuance is lost, and you can’t tell if the banding in the sky is a monitor artifact or really there in the image.
-
Maybe I work in different circles from yours, but I don’t know any professional colorists who work with 8-bit monitors for grading. Sure there are editors who do a bit of grading, but colorists? Not that I’ve seen. Viewing in 8-bit certainly does change a grade. To cite two simple ways it does, some dark nuance is lost, and you can’t tell if the banding in the sky is a monitor artifact or really there in the image.
As long as the underlying math is performed in 32 bit float, as in grading apps, the final output is not so critical. If you have recorded originals in 8bit, and there is banding in the sky, you will see it on an 8 bit display after grading. The risk here, and it's slight, is that there is some banding in an image on the 8bit output, but it is not visible in 10 bit output. For me, I have the ability to view 10 bits, but I usually leave it off to improve playback performance. It is very very rare for me to need to check a shot in 10 bits. And for all practical purposes, the 10 bit and 8 bit images display as a perceptual match.
There is no dark nuance lost in 8 bit output. Maybe in 8 bit originals, of course, but once graded in 32bit float, it's a moot point. It seems to me you are mixing together the grading math with the output pipeline. But they are two different issues.
When I work in Photoshop, I don't have the ability to view in 10 bits on my display as it's not an option on my GPU card, but I never see banding in the image that will disappear on 10 bit output. And actually, I don't think I can print or upload to the web 10 bit images anyways. My printer driver supports only 8 bit. So, a moot point again.
-
As long as the underlying math is performed in 32 bit float, as in grading apps, the final output is not so critical. If you have recorded originals in 8bit, and there is banding in the sky, you will see it on an 8 bit display after grading. The risk here, and it's slight, is that there is some banding in an image on the 8bit output, but it is not visible in 10 bit output. For me, I have the ability to view 10 bits, but I usually leave it off to improve playback performance. It is very very rare for me to need to check a shot in 10 bits. And for all practical purposes, the 10 bit and 8 bit images display as a perceptual match.
There is no dark nuance lost in 8 bit output. Maybe in 8 bit originals, of course, but once graded in 32bit float, it's a moot point. It seems to me you are mixing together the grading math with the output pipeline. But they are two different issues.
When I work in Photoshop, I don't have the ability to view in 10 bits on my display as it's not an option on my GPU card, but I never see banding in the image that will disappear on 10 bit output. And actually, I don't think I can print or upload to the web 10 bit images anyways. My printer driver supports only 8 bit. So, a moot point again.
I see banding on 8-bit display output that is not there when output to a 10-bit display frequently enough to know it’s a significant compromise for color grading work. We may simply be working with different sorts of images, and it may be less important to you.
-
I see banding on 8-bit display output that is not there when output to a 10-bit display frequently enough to know it’s a significant compromise for color grading work. We may simply be working with different sorts of images, and it may be less important to you.
Could be. Almost all my grading is done on movies that I've shot myself, and I'm always using professional level cameras that record 12 bit files, so banding isn't an issue that comes up often. When my grade goes so far as to break the image, it's pretty obvious on the scopes.
I just graded a short that I shot on a Sony A7RIII, 8 bit output to ProRes on an external recorder. There are some banding issues in this movie, but they are pretty obvious in 8 bit output, and they won't disappear in a 10 bit render. There were some drone shots, made with another camera, that had more issues that I did need to smooth over though... Of course the lesson here is to not shoot movies on a Sony A7RIII :)
-
I am on Windows using 8 bit output to NEC monitor which has 12-bit internal processing. After monitor calibration and with the use of the icc profiles there is absolutely no visible banding in color managed worklow, using Photoshop. The key here really is the 12-bit processing prior the 8-bit final output.
-
Could be. Almost all my grading is done on movies that I've shot myself, and I'm always using professional level cameras that record 12 bit files, so banding isn't an issue that comes up often. When my grade goes so far as to break the image, it's pretty obvious on the scopes.
I just graded a short that I shot on a Sony A7RIII, 8 bit output to ProRes on an external recorder. There are some banding issues in this movie, but they are pretty obvious in 8 bit output, and they won't disappear in a 10 bit render. There were some drone shots, made with another camera, that had more issues that I did need to smooth over though... Of course the lesson here is to not shoot movies on a Sony A7RIII :)
You're absolutely right about the A7 cameras. With limited color palets they can look very nice, but if you don't get the white balance right and want to bring it back, or you want to grade an exterior scene with sky, good luck.
When I was beginning to do color grading, a lot of my work was for the Maine Office of Tourism. We were shooting on Reds at very low compression ratios (because exteriors with water and trees can't stand much compression) and much of the fotage featured blue skies that were critical to the look. When I was sending the signal out a 10-bit pipeline from Resolve to an 8-bit monitor, I could easily break the sky with banding as I graded. When I switched to a 10-bit monitor, I found see that the grade wasn't breaking the sky at all, it was just the monitor that had been producig the banding. Proper monitoring gave me a lot more room to work those images. YMMV
-
When I was sending the signal out a 10-bit pipeline from Resolve to an 8-bit monitor, I could easily break the sky with banding as I graded. When I switched to a 10-bit monitor, I found see that the grade wasn't breaking the sky at all, it was just the monitor that had been producig the banding. Proper monitoring gave me a lot more room to work those images. YMMV
I might be getting around this issue by using a large 3d LUT for my display calibration. It seems possible to me, that using the .icc calibration of a display might more easily break the image when the output is 8 bits and the display is pushed into calibration in 8 bits. I say this, as years ago, when my display was not so good, I did have issues with banding after performing an OS level .icc calibration. Now, I'm using an Eizo display, which I think is much closer to the calibration target, before external calibration, and I'm not seeing issues sending it 8bit output in Resolve over the decklink video card.
Sure, if I send the display a greyscale ramp, if I zoom in closely, I can see steps in the 8bit output that smooth out when sending the display 10 bits. But on actual graded footage, this difference is almost always impossible to see on the display.
So, maybe this issue is more about the method of display calibration, combined with the quality of the display, rather than a strictly 8 bit vs 10 bit output issue.
-
I am on Windows using 8 bit output to NEC monitor which has 12-bit internal processing. After monitor calibration and with the use of the icc profiles there is absolutely no visible banding in color managed worklow, using Photoshop. The key here really is the 12-bit processing prior the 8-bit final output.
Michael, I think the issue here is the high bit processing you are using in Photoshop, rather than the 12 bit internal processing of the NEC display. Or... it could be both :)
When color grading video in a video grading app, you don't have access to the volumetric color data contained in the .icc profile stored in your OS and used by Photoshop. So, .icc/Spectraview calibration is not ideal for video grading. This issue has been discussed at length in the liftgammagain.com forum. The good news is, that if you are using Davinci Resolve, an iOne Display Pro probe, and a decklink video output card, then you can use open source Displaycal software to create a true 3D output LUT that you can use in Resolve. If you are adventurous, you can check out the Displaycal website to see how this is done. Or you can buy or rent Lightspace commercial software to do this as well, and it might be a slightly better calibration, but I'm not sure about that... But it's quite popular software for the professional grading business.
-
I will not turn this thread into a "how to calibrate" for video thread, but I just wanted to point out that many many professional colorists do work with 8 bit video output to their displays, at least for SDR/HDTV/cinema work and it's fine. Viewing in 8 bit does not change the grade, but maybe at the margins, helps one see any banding artifacts, or lack of them.
Errrr no.
Displays are 10 bit and the files are often 12 bit LOG / 16 bit lin.
No serious colourist is working on an 8 bit monitor. Calibration isn't going to change bit depth.
By the way, most BMD cameras shoot 16 bit LIN internally. They get encoded as a 12 bit log profile for recording to either ProRes 444 or to BRAW.
I suspect we'll see BMD video assists doing BRAW recording for cameras like the Nikons (and sigma FP and sony's) soon.
JB
-
Errrr no.
Displays are 10 bit and the files are often 12 bit LOG / 16 bit lin.
No serious colourist is working on an 8 bit monitor. Calibration isn't going to change bit depth.
By the way, most BMD cameras shoot 16 bit LIN internally. They get encoded as a 12 bit log profile for recording to either ProRes 444 or to BRAW.
I suspect we'll see BMD video assists doing BRAW recording for cameras like the Nikons (and sigma FP and sony's) soon.
JB
From my experience, viewing a graded video image in 8 bit SDR is ok. Ok, as long as the high bit originals are processed in 32 bit float from high bit depth originals. Banding in 8 bit, after grading is very very slight and usually hard to see. Because all the processing and rendering is done in high bit, it doesn't effect the delivery at all. Though there is certainly no harm in viewing in 10 bit, as long as it doesn't effect real time playback on your system.
That said, John is a very accomplished cinematographer who does beautiful work... So, please don't consider this an "argument" :) :) :)
All I mean to say is that if you can't have 10 bit display at the moment, it doesn't mean you can't accurately color correct your work, as long as the display is properly calibrated with a 3d LUT.