Magic Lantern enables 14-bit RAW CinemaDNG on 5D

i'm wondering if the vertical pattern on James Miller's latest might not be from the removal of his OLPF. I'm just not seeing anything like that on other videos. Not seeing jaggies of any kind. The highlights are the best thing about the hack as far as I can tell.
 
I hacked my 5dm3 and I'm not seeing any jagged edges or vertical lines- Im guessing the jagged edges come from raw processing in post (sharpening)- I wonder if the vertical FPN in some videos is from shooting at certain iso's? Its looking really good though- we just need to get this "embedded" color profile thing taken care of if possible, the default colors in ADR are way over saturated in certain channels.
 
hunter richards said:
I hacked my 5dm3 and I'm not seeing any jagged edges or vertical lines- Im guessing the jagged edges come from raw processing in post (sharpening)- I wonder if the vertical FPN in some videos is from shooting at certain iso's? Its looking really good though- we just need to get this "embedded" color profile thing taken care of if possible, the default colors in ADR are way over saturated in certain channels.
Click to expand...

Are people referring to the jagged edges around lights and the edges of buildings? I've seen a lot of them around street lamps etc in low light tests or whatever, they do seem less prominent in the H264 yet I would guess that's just how much it's being compressed.

Also curious where the ISO tests are? Those settings are clearly having an effect on the image.

Hunter, have you tried exporting TIFFs as is from ACR and then using an ACES workflow with them in R9? IDT CinemaDNG ODT none, see how that works?
 
I have used Tiff sequences directly exported from the ML DNGs in resolve 9 with input device transform with great success. My system plays back and colors well from the little I have done with it. If we could just get the DNGs playing as CDNGs. As of now they are recognized and playback but are not interpreted properly. It may be that resolve doesnt do 14bit natively. I think they are working on it.
 
dustwaterwindfire said:
I have used Tiff sequences directly exported from the ML DNGs in resolve 9 with input device transform with great success. My system plays back and colors well from the little I have done with it. If we could just get the DNGs playing as CDNGs. As of now they are recognized and playback but are not interpreted properly. It may be that resolve doesnt do 14bit natively. I think they are working on it.
Click to expand...

Thanks for confirming. I have a hunch that the "embedded" profile that we're discussing is actually how the RAW's supposed to be interpreted, similar to CDNG. You have to establish a LUT to view it say... flat. An ACES workflow similar to Ikonoskop, Blackmagic, and I think Sony F55 should be the way to work with this footage.

Agreed, the DNGs need to be converted to CDNG, though. TiFFs... I mean not bad, could be worse, but it's not really the way I'd choose.

I'm curious as to how Resolve's working with 14 and 16-bit, and if it does? Supposedly the 12-bit LOG CNDGs from BV1 are interpreted as 16-bit Linear in both ACR and R9, so it must already work with it no?

Really looking forward to R10 as well.
 
Kholi said:
Are people referring to the jagged edges around lights and the edges of buildings? I've seen a lot of them around street lamps etc in low light tests or whatever, they do seem less prominent in the H264 yet I would guess that's just how much it's being compressed.
Click to expand...

I won't be surprised if it turns out the reason 5dmk3's stock video is soft is exactly the need to cover for the aliasing and the fixed noise we see in the raw video source.
 
Ian-T said:
Yes I'm seeing lots of it in people's videos. Not only the vertical patterns but also the jagged edges.

I think this hack is cool and everything but if it's the feed to the live view that they are capturing then we should expect compromises in the final image. My first thought is what we are seeing was meant only to view on a small screen. On a small screen (viewfinder) everything we are seeing will look good. But to use this as video....well...expect there to be many issues with it. So I'm kind of not suprised about the stuttery motion, jagged edges, aliasing and fixed pattern noise.
Click to expand...

See any jaggies, FPN, or stuttering in mine? I don't see anything on my 24". The FPN is normal and can be cleaned up pretty well in ACR. There was a judder fix on the 18/5.
 
hunter richards said:
I hacked my 5dm3 and I'm not seeing any jagged edges or vertical lines- Im guessing the jagged edges come from raw processing in post (sharpening)- I wonder if the vertical FPN in some videos is from shooting at certain iso's? Its looking really good though- we just need to get this "embedded" color profile thing taken care of if possible, the default colors in ADR are way over saturated in certain channels.
Click to expand...

I heard visioncolor are creating a profile for ACR. I concur on the jagged edges most likely coming from oversharpening, I haven't moved from the default setting much and it looks quite analogue.
 
henryolonga said:
I agree. I am mainly an audio guy and the thing I treasure the most is reliability and I pay for it through my nose. Stuff has just gotta work 24/7 365 days otherwise I get mad.
Click to expand...

Same here but so far it's been rock solid. If it keeps performing as well as it has been I'm gonna start shooting some narrative tests next week and if that goes well I'll shoot my feature with it.
 
Samuel H said:
I'm seeing absolutely no color issues, whatsoever. In fact, I never saw such an accurate palette out-of-the-box!
http://www.dvxuser.com/V6/showthread.php?307967-Canon-RAW-video-color-accuracy
(I doubt it, but it could be a 5D2-vs-5D3 difference)
Click to expand...

There was a magenta tint issue with the MK3 build but it was fixed a couple of days ago. I've been flattening the image in ACR transcoding to ProRes 4444 and applying the cinestyle LUT in film convert, the magic hour video was processed with a Fuji stock. ProRes 422 HQ doesn't do the raw files justice. ML are working on CDNG, a bit of lossless compression would help too.
 
Last edited:
cpc said:
I won't be surprised if it turns out the reason 5dmk3's stock video is soft is exactly the need to cover for the aliasing and the fixed noise we see in the raw video source.
Click to expand...

Um, no. Not seeing any aliasing on the raw Mk3 vids. Just the FPN on a couple out of the hundreds that now exist, notably that of James a couple of days ago. But again, he's using a camera with the OLPF removed, possibly affecting things.
 
Cosimo Bullo said:
Um, no. Not seeing any aliasing on the raw Mk3 vids. Just the FPN on a couple out of the hundreds that now exist, notably that of James a couple of days ago. But again, he's using a camera with the OLPF removed, possibly affecting things.
Click to expand...

I've seen aliasing in every video, it becomes more apparent the more sharpening's applied and people are going overboard. Chroma MOire no, but there's aliasing. Not every hard to spot.

https://vimeo.com/66228674
 
Kholi said:
I've seen aliasing in every video, it becomes more apparent the more sharpening's applied and people are going overboard. Chroma MOire no, but there's aliasing. Not every hard to spot.
Click to expand...

LOL OMG there's aliasing! C'mon where's the aliasing in the magic hour clip I shot? I haven't even tried to optimise the sharpness settings yet. That Saad dude has got all sorts of problems with his footage, motion is all stuttery, I don't know what the hell happened there but it's not ML doing it. It's digital so sure there's probably a tiny bit of aliasing but it's nothing compared to almost EVERY other digital camera I've seen/used. To my eyes the only digital cameras that look more analogue are the Alexa and the F65.
 
squig said:
LOL OMG there's aliasing! C'mon where's the aliasing in the magic hour clip I shot? I haven't even tried to optimise the sharpness settings yet. That Saad dude has got all sorts of problems with his footage, motion is all stuttery, I don't know what the hell happened there but it's not ML doing it. It's digital so sure there's probably a tiny bit of aliasing but it's nothing compared to almost EVERY other digital camera I've seen/used. To my eyes the only digital cameras that look more analogue are the Alexa and the F65.
Click to expand...

So we basically agree, there's aliasing. Cosimo said he's not seeing any, I said it's there.

Where's the discrepancy? I must've missed it yet again =T

Also, I'd be interested if you could tell the difference between Alexa and another camera if they were 50/50 on the same project, on the web (which is where most people here have ingested Alexa footage, not RAW personally shot material). Care to take the challenge?

As much as I harp about mixing cameras, let's be realistic... all things considered equal, most of us couldn't tell.
 
You must log in or register to reply here.
Back
Top