Blackmagic Forum

I have, and it’s really nice.
I’m just looking for something more non-setup I guess?

And just more stuff to play around with. Anamorphic, handheld, etc

Nicolas from Autokroma wrote:With the BRAW API, before the 2.0 version (adding support for the new 12K footages), you could get from your BRAW 4 resolutions : 1/1, 1/2, 1/4 and 1/8. It is important because if you choose 1/3 or 1/16 resolution for playback for example, it is not optimised (If 1/3 : 1/2 will be rendered and then the software (Resolve/PPro) will downscale it : so it takes more time than if 1/3 was directly generated from the BRAW API).

On the new 12K footages, these resolutions are 1/1, 2/3, 1/3, 1/6. So the minimum resolution you can get from the BRAW API on a 12K footage is 2048x1080 . So the same as if you were with a 4K footage and 1/2 Resolution.

We hope to get new resolutions choice from the BMD API, especially for those new BRAW 12K footages, to improve playback performances.

I tested in Premiere Pro OSX with :

• with the official Blackmagic RAW 2.0 Beta1 : https://www.blackmagicdesign.com/fr/support/
• with BRAW Studio 2.2 Beta1 : https://www.autokroma.com/blog/BETA-For ... Science-V5

Tested in Premiere Pro OSX (with the official Blackmagic RAW 2.0 Beta1, but also ), 12K takes time to render (like at least one second for one frame in Full Res, rendered on CPU only with my macbook from 2012 so not a good workstation but that was just for a test), so decreasing the resolution playback is needed if you are not using proxies, and 1/6 is not enough IMHO !

Yes, that is what I was asking about. You really are rendering a 2k mpeg or something image (actually this sounds like something I wrote years ago. Pretty irrelevant anyway, as that is how wavelet codecs worked anyway) quickly on those laptops, like a proxy, not a 12k image. So, what is the real 12k recode time of changes. Are we going to be waiting at least 60 hours an hour of footage, to do a final version on some old laptop. How fast on a good laptop?

Did you read my entry about speed on a Mac mini with an eGPU?
If you can tell me precisely what you'd like to have rendered, like which of the demo clips, which effect at which resolution for output, I can test it.

BTW, the backlit shot of the 'angry young man' is showing chroma noise and minimal FPN only if I increase contrast in the dark parts to insane levels. Remember, it's a beta camera…

Thanks Uli

As a speed test, it's 12k, to see how slow outputting to a finale file is, graded however, it's mainly testing codec performance, so aslong as it thinks it has to recompress everything. We could probably extrapolate from the mini gpu's performance from there. I suspect it will be 10x+ faster than the laptop CPU. But is it possible to save to Braw yet?

I wonder if these systems do while you are still working behind the scenes frame processing yet, which would skew results too.

You can't save to BRAW, other than trimming. And you'd need to be more specific about grading, since that could be very simple, like one node, or very complex, like several nodes plus NR and more. The codec to save into doesn't need to slow things down significantly, one could save into ProRes instead of H.265.

Will it even be possible to render at 12K considering the GPU memory requirements? I suppose an integrated CPU/GPU would work since it can always use system memory (with a significant performance impact), but with a dedicated GPU we're already struggling at 4K sometimes. Throw some speed warp and noise reduction on there and watch it burn.

I know 12K won't usually be the delivery format. But what if you need to render out something at full resolution temporarily for VFX?

There are 48GB vram GPUs available and we have many projects rendered above 12k. Amusement rides for example, also in stereo3D.

Kim Janson wrote:When, where could we see those?

Peter Chamberlain wrote:There are 48GB vram GPUs available and we have many projects rendered above 12k. Amusement rides for example, also in stereo3D.

It would be theme parks. So Disney and the like.


JB

All time best post, John! Kim walked into that one.

This guy has done an in-depth analysis of the ursa 12k and whyiencea need to have camera. He also talks about the color science. some of it goes beyond me. But very ineteresting

http://filmscientist.com/mblog/2020/07/ ... necessary/

link Corrected thanks to john Brawley

Ricardo Marty

That link is dead.

Just truncate the link after "mblog", you'll get there.

This guy won't win any awards for coloring grading. As always, the 'product' is only as good as the weakest link, the camera can get lost in it.

Correct link

http://filmscientist.com/mblog/2020/07/ ... necessary/

JB

Another observation, for those without a Mac: my Mac mini with an RX 580 eGPU with (humble, at todays standards) 8 GB is handling the 12K demo clips even in a 8K DCI timeline. It does not crash. No GPU memory full message!
I can use temporal NR without crashing. I can use Aperture Diffraction, another demanding effect, without crashing. Of course, it's crawling at snail speed, but it's stable. It stalls with both added, but no crash.

I'm nothing less than impressed.

Kim Janson wrote:When, where could we see those?

Peter Chamberlain wrote:There are 48GB vram GPUs available and we have many projects rendered above 12k. Amusement rides for example, also in stereo3D.

Yes, when can we see the 16k-32k fullframe+ cameras Peter?

Uli, it would be a simple test of raw performance.

It's 12,288/6 not 12,288/3.



This is asswards as well.



Science.

Good Luck

Some rendering on above mentioned hardware:

4K DCI timeline, four nodes, one with temporal NR.
Rendered into 4K DCI ProRes 422 HQ, scaling and debayering (we'll need a new expression here) forced to max.

8K R3D source needs 0.6 sec/frame. 12K BRAW needs 0.3 sec/frame. Nuff said.

Uli Plank wrote:debayering (we'll need a new expression here)

-> demosaicing is the correct expression I learned from Hook.

well, i bit the bullet and preordered!
Super excited about this camera. hopefully it will arrive this month! (wishful thinking? haha)

I plan on moving to PL mount lenses with the SLR Magic APO Hyperprimes. They look great and i'm excited to try them out.

roger.magnusson wrote:Will it even be possible to render at 12K considering the GPU memory requirements? I suppose an integrated CPU/GPU would work since it can always use system memory (with a significant performance impact), but with a dedicated GPU we're already struggling at 4K sometimes. Throw some speed warp and noise reduction on there and watch it burn.

I know 12K won't usually be the delivery format. But what if you need to render out something at full resolution temporarily for VFX?

With AMD Metal or OpenCL (and maybe even with a NVidia card running OpenCL, haven't tested it) it's no problem. CUDA is fast but also very memory hungry.

Adam, very excited for you! Nice to know my APO primes will get a workout on that camera. Really looking forward to your findings. Having some test results with people in addition to everything else will be sweet eye candy.

Uli Plank wrote:Some rendering on above mentioned hardware:

4K DCI timeline, four nodes, one with temporal NR.
Rendered into 4K DCI ProRes 422 HQ, scaling and debayering (we'll need a new expression here) forced to max.

8K R3D source needs 0.6 sec/frame. 12K BRAW needs 0.3 sec/frame. Nuff said.

As I suspected. Thanks Uli. Still need a reasonable system.

I was going to update a thread I started with a link about what it was talking about. But somebody removed it for no reason.

But here is a link you guys might enjoy:

viewtopic.php?f=14&t=119014

Excited to see some 12K Ursa footage/trailer on YT BMD channel.. even though it was streamed full HD only.. I instantly was drawn to the capturing of motion, which looked very clear and clean.. fantastic..

but now seeing these readout times. I wonder how much of the 12k "trailer" used any actual Full 12k recorded footage?


Full 12K : ~15.5ms
12K 2.4:1 : ~12.3ms
8K/4K (full field of view) : ~8.5ms*
6K crop : ~7.8ms
8K 2.4:1 (full width) : ~6.7ms*
4K crop : ~4.25ms*

Camera would still be interesting as a 12k sensor outputting to 4k/8k with clean and clear motion.. but I always feel a bit jilted getting stills that are taken from a 12k sample.. and then seeing a trailer of a 12k camera not actually showing footage recorded in Full 12k.

Blackmagic RAW 2.0 Public Beta 2 is now live on the website with Resolve 16.3 Public Beta 2 and adds support for Windows

https://www.blackmagicdesign.com/support/

CaptainHook wrote:Blackmagic RAW 2.0 Public Beta 2 is now live on the website with Resolve 16.3 Public Beta 2 and adds support for Windows

https://www.blackmagicdesign.com/support/

Nice!

CaptainHook wrote:Blackmagic RAW 2.0 Public Beta 2 is now live on the website with Resolve 16.3 Public Beta 2 and adds support for Windows

https://www.blackmagicdesign.com/support/

Is that our 2020 resolve?
No resolve 17?

Come on, not bad for the Covid-19 year.

Since DR 16.3 is out, I think Somone will get URSA 12K please share High ISO thank you.

can anyone validate this as authentic Ursa Mini 12k footage?

visalapol wrote:Since DR 16.3 is out, I think Somone will get URSA 12K please share High ISO thank you.

80MP with crop sensor low light performance will be bad.

That is pretty cool. Like halfway between an Arri and an Red hydrogen in look. Notice the shine on the girls face?

Yogendra Singh wrote:80MP with crop sensor low light performance will be bad.

You’re using traditional bayer logic to draw a conclusion.

This is not a bayer sensor.

The low light performance is similar to a G2. So 1600 and 3200 are feasible. With the extra resolution, NR can also be more aggressive so you may be able to go further.

JB

Wayne Steven wrote:That is pretty cool. Like halfway between an Arri and an Red hydrogen in look. Notice the shine on the girls face?

That shine, is that good for a camera?

John Brawley wrote:This is not a bayer sensor.

Best statement in this thread.

Warning: Your preconceived notions may not be correct.

CaptainHook wrote:Blackmagic RAW 2.0 Public Beta 2 is now live on the website with Resolve 16.3 Public Beta 2 and adds support for Windows

https://www.blackmagicdesign.com/support/

but can't also not playing the 12k braw files on a cMP : -(

John Brawley wrote:
You’re using traditional bayer logic to draw a conclusion.

This is not a bayer sensor.

The low light performance is similar to a G2. So 1600 and 3200 are feasible. With the extra resolution, NR can also be more aggressive so you may be able to go further.

JB

Thats a good news.

Warning: Your preconceived notions may not be correct.

yea, surprisingly

Yogendra Singh wrote:

Wayne Steven wrote:That is pretty cool. Like halfway between an Arri and an Red hydrogen in look. Notice the shine on the girls face?

That shine, is that good for a camera?

Nope! But it turns up here in camera samples. It's likely what she's is wearing, or not wearing. Hence the need for 3 more stops, to make it look more similar to what we see, and more if you want to push it around and have a similar look to how we see. 16 stop should be minimum now. I was surprised G2 didn't get there. By what this camera does, a 4.6k should.

Just downloaded the newest version of Resolve which includes Gen 5 color for windows, however I don't see the Gen 5 extended LUT in resolve or in CST. Am I missing something

There is no LUT or CST support yet, Gen 5 is only through the Blackmagic RAW panel.
Blackmagic RAW 2.0 and Gen 5 are still in beta so you won't see LUTs or CST/RCM support until the GM.

Wayne Steven wrote:
Nope! But it turns up here in camera samples. It's likely what she's is wearing, or not wearing. Hence the need for 3 more stops, to make it look more similar to what we see, and more if you want to push it around and have a similar look to how we see. 16 stop should be minimum now. I was surprised G2 didn't get there. By what this camera does, a 4.6k should.

These highlight spots are what made Nikon to lose their market share to canon. surprisingly canon cameras with only 12 stops of DR have minimal highlight spots. even with 72 dpi pics.
I am sure BM cameras have better highlight roll-off.

CaptainHook wrote:There is no LUT or CST support yet, Gen 5 is only through the Blackmagic RAW panel.
Blackmagic RAW 2.0 and Gen 5 are still in beta so you won't see LUTs or CST/RCM support until the GM.

Thanks!

Hey everyone!

Long time lurker, first time posting.

I am admittedly still using and liking my Original BMCC 2.5k, but the new Ursa Mini 12K is very intriguing.

The footage from the new Ursa Mini 12K just looks very good. I am having a difficult time even defining it. I think it's the color, the rolloff and just something non-harsh that I often attribute to a lot of digital footage. Despite the resolution, for lack of a better way to say it, there is a flattering smoothness to it. I can only chalk this up to Blackmagic's new sensor design. I know a few people like John Brawley on his wordpress page have delved into this non bayer sensor a bit. I suppose I am trying to wrap my mind around this "new tech". I'm not sure it will even be fair to measure it using the previous standards or parameters.

As others get their hands on the camera, I would love to see more footage and hear their experiences with what they are discovering in post, as far as dynamic range, color, rolloff, etc. I also look forward to seeing how different lenses play with the new 12K. For instance the new Laowa OOOM 25-100mm Cine lens https://laowacine.com would be interesting to see. At just $5,000 for a Cine zoom, it seems like it might be a nice budget paring for the Ursa Mini. Or the new cine zoom lenses from DZOFilm look interesting as well. If anyone is able to get their hands on any of these new budget lenses, please share your results. I look forward to seeing and hearing about everyone's experiences.

Actually, Bayer was chosen because it was better at resolutions more one on one. There is few better ones, such as the random sensor pattern. Actually, it was originally meant to be complimentary coloured as well, but it the calculations needed were too much at that stage. But here, we are dealing with significant downresing images instead, allowing other patterns to be more useful. However, of rather the random sensor pattern, which isn't random at all, but enough to break up problems in demosaicing Bayer. I would rather a hybrid colour scheme instead too, using a few more complimentary colours. But seriously, three chip 4:4:4 like colour would be nice. The modern technology could suppress noise in the red channel too, and bring more low light. So, the sensor pattern isn't too much of a revolution, but it is a significant step towards the future.

Yogendra Singh wrote:

Wayne Steven wrote:
Nope! But it turns up here in camera samples. It's likely what she's is wearing, or not wearing. Hence the need for 3 more stops, to make it look more similar to what we see, and more if you want to push it around and have a similar look to how we see. 16 stop should be minimum now. I was surprised G2 didn't get there. By what this camera does, a 4.6k should.

These highlight spots are what made Nikon to lose their market share to canon. surprisingly canon cameras with only 12 stops of DR have minimal highlight spots. even with 72 dpi pics.
I am sure BM cameras have better highlight roll-off.

Without researching it, I will guess it's the following: The two sensor companies probably measure dynamic range in a different way, with the likely Sony technology not being able to keep up in the day, and doing something that raise the image to clipping I've commented about last decade or decade before last. Even now. The 686 Sony mobile sensor with its increased dynamic range, is still attracting review comments about dynamic range. However, dynamic range is only the start, it's what you do with it. As I was posting about a while ago, you can carefully clean the noise floor and restore some picture to get extra stops, use highlight recovery like techniques, and use the extra space to move the image response around in. Of course, it's always possible, that Cabin users being so stayed and conservative, under expose the image reducing it, and those hyped up Sony radicals are over exposing.

Of course, Sony products tend towards a brighter image, and clipping the low end to black (as can still be seen sometimes). In film days people didn't have such an issue with noise in shadows, without clipping it off. They didn't have trouble with noise everywhere, though fortunately not on the Sun. The usual type back then, were worried how bad the image looked with noise, and even tried to find ways to put to grain back in. Now days I am glad it is largely gone. Some noise is ok, and more noise when you don't have choices, or for effect. But, as usual, these types never come back and say you were right and apologize.

Adam Langdon wrote:can anyone validate this as authentic Ursa Mini 12k footage?

Based on their followup vid also titled, 12k URSA.. I'd say this is click-bait.

In that other vid, the reflection shows a 4.6k Ursa Mini, although it could be an early mock-body for the 12k... safe to assume this is just someone doing click-bait...effectively.

Has anyone tried pushing in on the 12K samples 2 or 4 times? I'm not getting good results.

The Film and Digital Times (FDT) Issue 104 features John Brawley and the URSA Mini Pro 12G:
https://www.fdtimes.com/issues/august-2020-edition/

Individual back issues may be freely downloaded as well. Enjoy.

Interestingly, they found the Sony A7S III worth covering too.