With the URSA Mini Pro 12K, Blackmagic introduces a new sensor that changes how we might work.
Blackmagic launched the URSA Mini Pro 12K earlier this year, and beyond simply changing the game in terms of resolution, leapfrogging from their current 6K Pocket right over 8K and going straight to 12K, the camera boasts a completely fresh sensor design, which allows the camera to operate in ways other cameras don't—at least not yet.
Building on its existing URSA Mini camera body design, the 12K version delivers such stunning images that you're able to consider the camera as a real contender when it comes to higher-end productions. Blackmagic Pocket cameras have been dominant, and now we're about to see the URSA Mini Pro on more productions than ever before.
The biggest news is the custom-designed sensor that delivers 12K resolution. Though we don't know what exactly went into the development of it, patent applications do shed some light on what is most likely the sensor pattern, along with information John Brawley has put out during his initial testings on the pre-production model.
We spoke with a representative from Blackmagic to find out more about the sensor, and they wouldn't say much beyond what was presented in the initial release. It's somewhat expected as with any trade secrets.
What we do know is the 79MP sensor took three years to make, and it's not a Bayer sensor. Let that set in. It doesn't use a Bayer pattern Color Filter Array (CFA). Instead, as Brawley previously pointed out, it uses a brand new custom filter array that has an equal number of red, green, and blue photosites, as well as clear or white photosites.
What's unique about that? Well, it's a digital camera sensor that doesn't see color unless there's a Color Filter Array added to it. If there isn't one, the sensor only generates a black and white image. Normally in a Bayer sensor, the color filter is added on top of the photosites where each grid of 2x2 pixels receives two green filters, a blue filter, and a red filter. That allows the sensor to create a color image. The processing of the data from the CFA is where camera manufacturers tend to customize the look of the color science. The color science that processes the bayer CFA data is partly why Canon, Sony, ARRI, or RED footage look a certain way straight out of the camera.
With the URSA Mini Pro 12K, instead of a Bayer 2x2 grid of Green-Red-Blue-Green (GRBG), it has a 6x6 grid. Meaning six green, six blue, and six red, plus 18 white. The sensor, in some ways, is like having two 6K sensors, on top of each other. It combines roughly 2K red, 2K blue, and 2K green pixels, which is necessary to create the color image, with roughly 6K white or clear pixels that let all the light through without going through a color filter.
The sensor also has a very small pixel pitch of 2.2. microns. One of the primary difficulties in increasing resolution is losing low light sensitivity. The smaller the photosite on the sensor, the more light you need to retain detail. By removing the color filter from half the pixels, Blackmagic is able to have a sensor that's still good at low light imaging. That is since some light is lost to the filters in the CFA; by having no filters on the clear pixels, less light is needed to generate a signal.
Since humans don't see a lot of different colors in shadow areas, and color casts in the shadows of a digital camera are generally something we want to avoid, this is a pretty brilliant compromise. It gives us high resolution combined with good low light reproduction.
The native ISO of the camera is ISO 800, and Blackmagic says it has 14 stops of dynamic range. That said, I don't trust any manufacturer's advertised dynamic range figure.
It generally is measuring the signal-to-noise ratio inside the camera at the sensor, or it's testing using something like a DSC Labs chart. While they might be scientifically accurate, I don't find them useful for how I shoot. Even testing dynamic range with Imatest, there's isn't a single way to test dynamic range. The parameters you decide to shoot in, e.g. ISO, gamma curve, may have a different dynamic range than the stated maximum.
When doing over/under tests with the URSA Mini Pro 12K, I was quite pleased with the latitude. It definitely has 12 or maybe 13 usable stops of latitude, with something like 5 or even 6 stops above key and at least 7 stops below key.
What was most impressive was how well it recovered from overexposure with the outdoor scenes. I could grade shots that were as much as 4 stops over to something that was acceptable though not ideal. This is absolutely astounding and definitely something to appreciate about the camera. Latitude is of course something that can be controlled with lighting on set, but having that massive range of exposure will be super useful in bright exterior situations or highly contrasted night scenes.
Less surprising, but still wonderful, is how much information was still present at 7 stops under. The sensor with its clear pixels is designed for that. Not that there is great information down at 7 stops under. It's noisy, of course, but there's picture info there, and that's useful. This will absolutely be a camera that is a pleasure to work with in low light.
It's worth noting, I did get some infrared light interference in the underexposed indoor shots when they were lifted. As you see in the red/pink cast in the graded shots, the camera lacks either an OLPF or an IR filter, which means if you are using a heavy IR source, you'll see some of this issue. This can be easily fixed on set when using an IR front of the lens, but we wonder if it's possible to get an after-market internal ND with better IR protection.
Whenever a new camera resolution comes out, there are always some nerds (myself among them) who like to remind users that the stated resolution, say 4K, isn't really 4K if it needs to be demosaiced. Once you demosaic, you end up with less actual working resolution. A 4K camera that gives you around 2.5K of resolution is giving you the same functional resolution as the 2.5K Alexa, which demosaics in-camera before creating the ProRes files.
That, of course, is still true with the URSA Mini Pro 12K. You don't have 12,000 red pixels, 12,000 blue pixels, and 12,000 green pixels, which would really make for a 12K RGB image. Instead, you have 6,000 white pixels, 2000 red, 2000 blue, and 2000 green, and then it's processed to interpolate back to the 12K RGB pixels of the file.
You know what?
12K is marketing, but marketing works. I know operators who own original Alexas, which still shoot beautiful images that have identical measurable resolution to a "4K" camera that shoots Bayer pattern images, but clients prefer the 4K. Sometimes clients don't always fully understand what they are asking for, but 4K has marketing power and those Alexa owners sometimes need to rent Sony FS7's for jobs they could easily shoot on the original Alexa, and it would be gorgeous.
12K will have real power with a lot of clients. Blackmagic is also pushing the HDR as well, which seems just as, if not more, important with this camera's healthy dynamic range.
The real surprise, one I will admit to have slept on when the camera was first announced, was the full sensor readouts. You can shoot 12K, 8K, 6K, or 4K all full sensor. This is a great feature of this camera. Maybe the defining feature.
I believe this is possible since symmetrical math is easier to process than debayering math. The camera is capable of taking that full 12k image, demosaicing it inside the camera, and creating a 4K Blackmagic Raw file inside the camera. All without windowing in on the sensor, which is what most competitors do. If you think about it, it makes sense. It's easier to take 12K and 6K from a symmetrical sensor (just get all your computations in half) than it is to get 4K from 8K on a Bayer sensor where the computations have to take the Bayer pattern into account, which is more complicated mathematically.
Why should you care?
Well, by shooting with the full sensor, all your lenses shoot the field of view you are used to, and you have a functional set of coverage. Your 18mm looks like an 18mm. You don't suddenly need an 8mm lens to get a wide shot because you've windowed down to 2K resolution. The biggest frustration when shooting in slow-motion on other cameras is that we need to take out extra wide lenses to get back to fields of view when you window down the sensor.
Besides slow-motion, is there any other reason to shoot 4K on a 12K sensor? Well, those 12K files are beautiful, but for many jobs, they might just be overkill. (For example, shooting a documentary that has a tight turnaround.) There's a strong argument that 6K, or even 4K, is plenty of resolution for a television finish, while 8K or 12K might be overkill. The beauty here is that if you want to shoot 4K, it's real full sensor 4K. Not cropped 4K. The 4K images that it does create are beautiful and are often indistinguishable from 12K images on a laptop or even a 50" monitor screen. Those 4K files will be much smaller on your hard drive, much easier to edit, and turn around faster than the 12K files.
In fact, for documentary work, this is likely going to be mainly a 4K or 6K camera. For independent features, where you might want more reframing options and might be willing to splurge more on storage space, you'll want to consider 8K or 12K. Same with commercials and music videos. But of course, there always are exceptions, where a doc might have a reason to shoot 12K, or you might be shooting a heavily improvised feature that makes more sense in 4k for the hours and hours you'll be shooting. But the key is that you have the choice without having to get a whole separate sense of lenses out.
Blackmagic released a firmware update that added several new frame rate options to URSA Mini Pro 12K.
- 12K 2.4:1 maximum frame rate to 75fps
- 8K DCI, 16:9 and 6:5 Anamorphic to 120fps
- 4K DCI, 16:9 and 6:5 Anamorphic to 120fps
- 8K 2.4:1 and 4K 2.4:1 to 160fps
- 4K Super 16 to 240fps
The camera can do 4K 120fps in full sensor, 2.4:1 4k to 160fps, or if you want 240fps, which is the realm of real slow-motion seriousness, you can go down to a 4K Super 16. The Super 16 is actually really pleasing imagery and didn't feel anywhere near as "artifacty" as windowed views sometimes feel with other sensors.
When you crop in too much on other Bayer sensors, the artifacts start to really pile up, even on the RAW file. But cropping in on the symmetrical sensor layout of the 12K, it felt kind of nice. I would be tempted to shoot a whole project in that mode, or even to invest in some affordable Super 16 primes like the original Zeiss Super Speeds.
The slow-motion footage above was shot in 4K UHD Super 16 at 240fps.
While the camera supports SSD, CFast, and SD cards, for my money, this is really just an SSD camera. CFast is overpriced as storage as it's roughly 1/3 the price of RED storage, but it's still eight times pricer than something like NVMe, which means RED storage is something like 24 times the price of NVMe. Since this camera also supports NVMe based USB drives it's the only thing that makes financial sense.
It's $220 or so for 1TB of NVMe versus $250 for 128GB of CFast. There is just no comparison whatsoever. While a nice option, Blackmagic could almost drop the CFast card slots from the system. I did one shot on the CFast card, and it just filled up so fast it was ridiculous. I ended up mainly using Caldigit Tuff Nano and Samsung T5 drives.
One idea for an accessory would be a drive holder that screws into the rosette on the righthand side. Some sort of clip or mount would go a long way. That said, I used velcro while shooting and it worked just fine. There is a little screw receptacle next to the USB port to lock in the USB and prevent accidentally knocking it out. Hopefully, third-party manufacturers will launch robust USB-C cables soon with the built-in screw to lock it in place. USB does make me a little nervous as a field connector, and some sort of locking USB-C would be wonderful.
The perk of NVMe is that it's just a USB drive. Buy enough for your shoot, shoot to them, and when they are full, your DIT can make backups, and away you go. It's cheap enough you can just work off the NVMe drives themselves, or store them in a closet and not reformat them to shoot more.
One of the nice features of the camera is that it is the same basic design as other URSA Mini Pro cameras that have been out for years now. This means that vintage accessories will work. You don't need to buy a new pistol grip, shoulder mount, or viewfinder.
For accessory cost-saving, legacy hardware is very appreciated despite the truly innovative sensor design. However, one place where the body is starting to feel a bit long in the tooth is the SD and CFast card slots. They don't seem likely to be used that often, and it seems like a refresh of the body design should probably come out sometime soon that deletes that area and opens up either CFexpress or perhaps an NVMe slot.
For future generations, Blackmagic could shorten the overall body length of the camera, which is one area where they are different than the competition. This body is large enough to require a shoulder mount, and the full-sized Ronin model, but something shorter would be a game-changer for rigging.
12K is a lot of resolution. However, it is mosaiced 12K, so it's really probably something like 8K after demosaicing. While there are very few lenses that truly resolve 8K, I shot some tests with a variety of XEEN CF, Sigma Cine Prime, and IRIX lenses and was quite impressed and pleased with the results.
If you're finishing 4K, and honestly 4K finish is probably all you need to do for the next few years, you can easily get away with 20-25% crop before any noticeable artifacting occurs.
So, should you shoot your wide shots assuming you'll zoom in for a close-up? No, of course not, shoot a wide shot and then a close-up. But if the boom dips into frame, if the horizon is off, or if you want a slight tweak, you have a pretty beefy amount of room to punch in if you need it.
All of the tests are finished in 8K. If you happen to have an 8K monitor, you can watch in full 8K. Though with today's compression limitations from streaming services, it may not look much better than a 4K image. It's far more likely that you'll see noise from the internal compression or the final stage delivery compression than you will see resolution artifacts from 12K. We're definitely in the realm of serious oversampling here.
Also worth noting is that rendering out on Resolve in 8K took much more than four times longer than rendering out of Resolve in 4K. Not sure precisely what is going on there, but a render that we could do in 4K in about 10 minutes would take hours in 8K. We had the same experience with Vimeo. 8K videos took way longer than 4K videos to convert for Vimeo playback. Significantly more than the four times as long than you might expect. If you're planning on delivering to 8K, plan on much longer renders and longer conversions.
One place 12K could really shine is for platforms where you need vertical and horizontal 16x9 images with a lot of resolution and room to reframe. For something like Ficto where you need both masters, 12K will have you covered. There are also likely going to be other immersive content experiences where the 12K might make sense. For instance, at a museum or for an art installation.
But most people won't have more than 4K at home for a long time. The same for theaters, once they're back open. You'll get more bang for your buck mastering in HDR, which is more noticeable a difference, than the jump in resolution.
Color & Look
To my eye, for the first time on a Blackmagic camera, the color reproduction was fantastic, rich, and pleasing.
It uses the new Blackmagic Color Science 5 and is a dramatic change from the previous generation URSA Mini Pro cameras. It manages to feel saturated naturally but not overly digitally saturated. Skin tones are pleasant without an orange or red cast that needs to be timed out. And there aren't dramatic color casts in the shadows which sometimes plague skin tones on some digital cameras.
Blackmagic has put in a lot of work making this footage look good right out of the box. The image you monitor, and the image you bring up in Resolve without any correction, just looks nice. The skin tones feel natural.
But of course, with Resolve you can do a ton to manipulate the color grade. Despite the power of final grading, reputations sometimes get made or broken based on how dailies look. If the director and editor spend a week looking at unpleasant dailies while they edit, even if it looks amazing at the end, the credit might go to the colorist rather than the DP.
I know some DPs who lost clients over dailies that didn't look amazing. Crazy.
Blackmagic was smart to put the work into making a file that comes out of the camera looking nice.
In the tests, I used Extended Video as our main viewing LUT in-camera, which very useful but it has some extra contrast that doesn't show off the full range of the sensor. If we had just trusted that LUT, we would think the camera had more like 9 stops of latitude, which makes sense in terms of its purpose, which is helping make on-set decisions.
In grading in Resolve, I didn't use a LUT, and simply turned the contrast up to 1.65 in our primary node. Then I used the exposure tools in the RAW tab to tweak exposure to compensate for over/under. You should feel safe that there is some extra room beyond what you see with the Extended Video LUT on set.
The camera only shoots Blackmagic RAW. No ProRes, no DNxHD, no CineForm. Just Blackmagic RAW.
This is a rare situation for a camera in 2020 but it's understandable as to why. It is likely a result of the processing required for the symmetrical 12K sensor that they weren't able to support other formats. In addition, the software needed to handle BRAW is widely available. Resolve is free, the camera comes with a license for Resolve studio, and there are Blackmagic RAW plugins for Premiere Pro and Media Composer.
One option that some could consider would be using an external recorder like the Atomos Shogun or the Blackmagic Video Assist to record ProRes files. This would allow you to record proxy files in DNxHD or ProRes that would be faster to edit in non-Resolve workflows.
One frustration is that there isn't a smart link between the camera and the Video Assist, despite being both Blackmagic products. It would be great if there were something that would allow for synced start/stop, so the RAW files in the camera and the external ProRes files get the same timecode and shot numbers. This could be something added in a future firmware update.
There's a detailed article devoted to the post workflow, but I wanted to mention a little bit about it here. 12K RAW files are processor intensive. They are going to cut and transcode more easily on a more powerful system. 12K raw is going to work well on the new Mac Pro towers that came out in 2019, the newest MacBook Pro 16" with the best graphics card, and other similar workstations.
Working on a smaller or older laptop, you are going to struggle, and we recommend making proxy files to make working with them easier. It's a tried and true workflow that you should practice regardless since cameras seem to always stay four to five years ahead of computers in terms of their power.
Some early users have reported working natively in 12K raw on their laptop, but if you don't have the newest MacBook Pro with the extra GPU upgrade, proxy files are the way to go. We were able to play a single stream of 12K RAW on a 13" Macbook Pro just fine, but as soon as we started staking up layers, or doing cross dissolves or other editing work, we saw some lag. If your edit station is underpowered, proxies are the way to go.
One thing to especially consider the Blackmagic eGPU. Availability is pretty limited and it's mostly found on resale sites, but my testing found that it could speed up render times by as much as 500% when working on a less powerful system like a 13" MacBook Pro. This was a complete game-changer for working with 12K files on a weaker laptop.
We suspect Blackmagic will come out with a new eGPU at some point in the future, but the used options available are very attractive. The eGPU doesn't speed everything up, but when rendering optimized media or dailies in Resolve, you can see the benefit.
My reservations with the camera are very limited. The images are frankly so good and are such an evolution for Blackmagic that this camera feels like a viable option in the cinema space when comparing it to RED, ARRI, or Sony VENICE. This is definitely a step up to the plate for Blackmagic.
It's not likely going to take over the top of the studio system where ARRI thrives, but there are some indie projects that should seriously give this camera a look. Indie features usually can afford the post workflow time and the hard drive space to make 12K worth it.
There also might be some argument for mixing resolutions as some indie films did back in the film days, cutting together 35mm and Super16mm. I could see an argument for shooting wide and establishing shots in 12K, then going into coverage in 6K for the long dialogue shots where the resolution might not be needed, but you'll be shooting a lot of volume.
To me, it seems like this is a camera that will be mostly shot to external NVMe SSDs. That said, it would be nice to see the SD and CFast slots removed and replaced with something else or less space. It'll also be amazing to see this in a full-frame sensor format down the road, though that would be something like 16 or 18K, which feels ridiculous.
The other major reservation is thinking about how soon it will be until there is a 12K Pocket Cinema Camera. The Pocket is a huge hit for Blackmagic, and frankly, a 12K Pocket will be a monster seller, even if it comes in the $3-4K price point. What you'll lose, of course, will be pretty significant. The beauty of the URSA Mini Pro body are the robust connectors and buttons that make it a truly usable camera for life on set.
You can use Gold or V mount batteries, power your accessories, mount an internal SSD, connect full-sized XLR connectors from sound, patch out various SDI options, and access all your menus quickly and easily while you work. The Pocket won't allow that, and will likely limit some of your slow-motion options just due to processor overhead.
The Pocket will also have a single lens mount, which is limiting. We enjoyed shooting PL and EF mount glass on the URSA Mini Pro, and to get that flexibility in the Pocket, Blackmagic will need to go with a shallow mount like RF, or the L mount. If they stick with EF, you can adapt easily to PL, but that still leaves out a lot of lenses you might want to consider.
Make no mistake, there probably will be a 12K Pocket, either in 2021 or in 2022, so your decision to invest in the 12K now should be made knowing that.
The other reservation that will be a deal-breaker for some is the RAW-only recording, which is something unlikely to change with firmware. If you are old enough to have worked with the original RED One back in the day, this is a familiar setup. Shoot RAW on set, transcode to something workable, and reconnect to RAW for color grading. However, in the last decade many have gotten used to shooting straight to ProRes or DNxHD for faster post workflows. Not having this feature is something that will be a deal-breaker for some.
There is also the worry about moire, which didn't bother me nearly as much as I thought, even though the camera lacks an OLPF filter. That red tartan jacket is in the test to specifically kick up moire and IR. You definitely see IR in the underexposed shots of the over/under test, though this would be easily fixed with a hot mirror filter. It is a frustration that there isn't something like that already built in. Perhaps an aftermarket upgrade to the internal ND is in the works.
This kind of game changer is what we used to expect out of RED. It's a totally new sensor design that simultaneously gives you incredibly pleasing images with a wide dynamic range and great color. It also makes the camera more flexible for a variety of jobs. Its limitations are small if you are used to working with a RAW workflow, and perhaps there is a way for Blackmagic to develop the Video Assist line to provide a viable realtime dailies solution.
Blackmagic has definitely stepped up their game, and have a whole ecosystem working together in a way that few could've imagined a few years ago.