Magic Lantern enables 14-bit RAW CinemaDNG on 5D

For those asking about the 5D2: AFAIK, RAW recording will come to that camera too, but they're working more heavily on the 5D3 now. In any case, the 5D2 will have aliasing and moire issues in RAW video, which the 5D3 doesn't have.
 
I wonder what the quality will be like on the 5D2 and if it will have comparable results to the 5D2 and if it will be fast enough to get similar recording specs.
 
That is why people say Alexa "like". Hence "like". Do you really expect a HDslr hacked to perform like a $90,000 camera? Or can you live with a "like" like image? I'll say this. The BMC is considered Alexa like. Minus all the rainbow moire that is :2vrolijk_08:

Personally yeah. I don't have any other choice at my budgets. But I'm still curious as all hell what it is they are doing on the Alexa to make objects on screen move like that.

As far as I'm concerned the lesser cameras now have everything else down when it comes to film-like. Or close enough anyway. But that subject motion is the one remaining missing piece. Yet it's the one piece that is the suspension of disbelief for me personally.
 
Personally yeah. I don't have any other choice at my budgets. But I'm still curious as all hell what it is they are doing on the Alexa to make objects on screen move like that.

As far as I'm concerned the lesser cameras now have everything else down when it comes to film-like. Or close enough anyway. But that subject motion is the one remaining missing piece. Yet it's the one piece that is the suspension of disbelief for me personally.


Really excellent processing, top notch engineering, best read out of the sensor possible. Go to their page and just look at each component, it's like looking at a Ferrari.
 
For those asking about the 5D2: AFAIK, RAW recording will come to that camera too, but they're working more heavily on the 5D3 now. In any case, the 5D2 will have aliasing and moire issues in RAW video, which the 5D3 doesn't have.

And it won't have the same performance for sure, also to evaluate for first findings with the same card:

"I can get up to 1880x720 no problem on Mark II with KomputerBay 1000x CF. Any higher than that and I start to drop frames."

http://www.magiclantern.fm/forum/index.php?topic=5247.msg35013#msg35013

vs

"On my 1000x card anything up to 87MB/s is pretty reliable, no skipped frames.

Full 3.5K lasts about 2 seconds before the buffer fills up and the card can't keep up.

Dropping to a still very respectable 2560 x 720 at 24fps seems to work well, no buffer overflows."

http://www.dvxuser.com/V6/showthrea...-on-5D&p=1986318367&highlight=#post1986318367
 
  • Like
Reactions: CBS
Woah.Smith rock is a knockout. Wide and deep. Keep em coming. I have a 5D MK II. I sure wish it could get close to this as most of this is MK III right. Squig - will it be ported to the MK II do you know? Looks truly, truly High end.
Best
Henry

There's already a build floating around for the MKII, not quite 1080p yet, something like 1880 pixels. I think a VAF filter will be an essential investment with the MKII because the aliasing is even more pronounced RAW. Detail and dynamic range wise it looks every bit as good as the MK3.
 
+1
That first shot with the pebble trail on the BMC is completely unusable. In fact, the Mk3 saved the day on that shot. I'd take a small hit on resolution any day for the confidence of having a moire free image.

The night street shoot I did was the true resolution test for me. Anamorphic softness, lens wide open, denonising, underexposure, and it's still film sharp with the sharpness slider at 50 in Adobe RAW. Even the H.264 transcode looks good. In controlled lighting you could never want for sharpness.
 
So; this does not work on the newest MIII firmware (the HDMI out one)? Does anyone know if ML will be working on a hack that works with the latest firmware? Otherwise, is it possible to roll back to an earlier firmware on the MkIII?

Nope not on the new firmware but they will try to port it eventually. Yeah you can roll back, It was a bit stubborn when I did it but it worked eventually.
 

LOL. I had to watch it a few times. Yeah I can see it on the doors. The vertical FPN has always bugged me on the Canons, what ISO was that? In my ISO body cap test 1250 ISO had the least FPN. I'll do a new RAW ISO test. Try the Adobe RAW denoiser, that fixed it in my night test but there's still a bit of a dirty foggy looking vertical smudge in the shadows on the pans.
 
Here is an Alexa video from the same people as the Smith Rock video. Compare the two and you can totally see the difference in subject motion. Even at the same frame rate.

There is a reason why cheaper cameras will always look "worse" than expensive cameras in motion. The operators. That Alexa flick you posted was stabilized with Gyros. How many DSLR user can do that? Films shot on Alexa are usually thought out a bit more. The shots are planned. Everything is done more properly. Movement is not just pointing the camera somewhere. Why do people have that misconception? If you want filmic motion, you need to move the camera filmicly.
 
Since I don't have a fast 1000x CF yet, I coaxed my poor 133x card to capture what it could just so I could get familiar with the workflow and the image.
Here's a video of things around my house, shiny utensils to test for highlight recovery, detailed stuff like flowers, etc. Mostly at 1280x720 upscaled to 1920x1080.
Towards the end are various res chart tests for those interested on the resolution,moire and aliasing as compared to native h.264 and also showing how the Mosaic filter still
is needed even with raw dng recording.


Vimeo link

 
Last edited:
Does anyone know if it is possible to use this new ML hack, recording RAW, but onto the Atomos Ninja 2? so you wouldn't have to buy a 1000x card and can just use an SSD?
 
Does anyone know if it is possible to use this new ML hack, recording RAW, but onto the Atomos Ninja 2? so you wouldn't have to buy a 1000x card and can just use an SSD?
Based on the technical exchanges I read between developers at ML forums, they say it wont be possible since HDMI is like 8bits only? (needs confirmation) compared to 14bits of the dng raw.
But, hey, months ago, ML devs also said raw wasnt possible, so who knows.
 
There is a reason why cheaper cameras will always look "worse" than expensive cameras in motion. The operators. That Alexa flick you posted was stabilized with Gyros. How many DSLR user can do that? Films shot on Alexa are usually thought out a bit more. The shots are planned. Everything is done more properly. Movement is not just pointing the camera somewhere. Why do people have that misconception? If you want filmic motion, you need to move the camera filmicly.

Concerning the Alexa. It has something most video cameras don't. A mechanical shutter. (And an optical viewfinder)
This means that the way the image is exposed is actually much closer to the way film is shot.
So in a sense it is neither rolling shutter or global shutter, but a bit of both.
This is a big part of why it does not have the same look as a conventional video camera.
 
Concerning the Alexa. It has something most video cameras don't. A mechanical shutter. (And an optical viewfinder)
This means that the way the image is exposed is actually much closer to the way film is shot.
So in a sense it is neither rolling shutter or global shutter, but a bit of both.
This is a big part of why it does not have the same look as a conventional video camera.

The basic Alexa that most people use has neither mechanical shutter nor optical finder. Only the top of the line studio model does. They do have a good OLPF and a built in low contrast filter that contribute to that creamy smoothness and nice gamma curve. Plus rolling shutter is twice as fast as any DSLR, comparable to film cameras.
 
Based on the technical exchanges I read between developers at ML forums, they say it wont be possible since HDMI is like 8bits only? (needs confirmation) compared to 14bits of the dng raw.
But, hey, months ago, ML devs also said raw wasnt possible, so who knows.

HDMI is not bit depth limited, it is only a pipe. Bit depth is limited by encoding formats. Under the current v1.4 deep color standard HDMI will support up to 14 bit video encoding. Ninja only records encoded video though. The raw dump from the Canon would just look like noise. From what I've read the raw dump to the SD card is not even formatted DNG. You have to run the ML utility on the data to get to a DNG formatted file that a raw image processors can recognize.
 
Anything is possible but ML are gonna focussed on internal RAW and porting it to other cameras for the foreseeable future. I think their prime focus is to get high quality reliable RAW recording functional on the 5D MK3, audio wasn't even an initial consideration. I think they've been dreaming of exactly the same camera I've been dreaming of: a small, lightweight, inexpensive digital cinema camera that shoots 35mm film like images, everything else like sound be damned. Anything else is a bonus.
 
Last edited:
Back
Top