Cinelog - True logspace conversion for DNG and CinemaDNG footage

[dubzeebass]

Hi Andy just bought a c100 mark ii. Can you do a c-log to cinelog or cineon to cinelog transfer possibly?

[Andy600]

@dubzeebass - Congrats 

Canon Log in bt.601 (for 1DC 4k), bt.709, Cinema gamut, DCI, DCI P3+, C100/C300 CPLock (using ACES IDT transform) and WDR10 input and output transforms are all in the coming update plus we can add others if the Mark II uses something different or has a firmware update to Canon log2.

[dubzeebass]

Thanks Andy! When's that update due out? You rock dude!

[clemento]

Hi Guys,
ok so I just purchased cinelog-c, but somehow I can't open the profile in ACR.
I put the "Canon DSLR for Magic Lantern Raw Video" folder (with the dcp profiles in it) into C:\Users\name\AppData\Roaming\Adobe\CameraRaw\CameraProfiles but when I open after effects and import a dng sequence I haven't got the cinelog option in the profile dropdown menu.
I've worked with visionlog and this profile is working well in ACR, but cinelog isn't showing up?!
My ACR Version is 8.1
I'm using dng material shot from the 5dmIII ml raw.

What am I doing wrong?
I've also installed the opencolorIO plugin.

Please help, thanks so much

[Andy600]

@clemento - If the profiles are installed into the correct location the UniqueCameraModel tag is the likely culprit. Can you send me a DNG and I'll check for you.

We recommend using .MLV instead of .raw and extracting/converting DNG from .mlv files with either raw2cdng (PC) or MLVFS (Mac) as they embed deeper metadata.

I can send you a custom profile for your incorrectly tagged dng files (they are likely tagged as 'Canikon'). Can you let me know your camera model?

[clemento]

Hi Andy,
thanks for your fast reply. Yes it says Canikon when I open the dng files in ACR.
I'm shooting with a 5dmIII and various lenses.
here is one of the dng http://we.tl/Quf6ubmKEF
it's 50fps and has to be stretched by 1.61.
I do have now a solid workflow solution with .raw files, creating proxys etc.
so if anyhow possible I would stick to raw-rec. Can I use your custom profile for future shoots with the same camera(5dmIII), also if iso, aperture etc. changes?
thank you

[DeafEyeJedi]

@clemento:

Could you provide us one good reason why you'd rather shoot RAW over MLV (other than it's deep metadata which helps a lot with colorgrade, etc)?

[clemento]

I changed my mind
I thought it would be difficult to generate dng's and proxys from the mlv format on windows, but I checked "phreekz" raw2cdng converter and this works brilliant
Not so sure what the "mlv_rec.tmp" file is which is generated each time on the card while recording. Is it some sort of  temporary cache?
Thanks

[Licaon_Kter]

It's the "Reserve card space" feature, it ML -> Movie-> RAW menu

[dubzeebass]

Today I discovered that the footage from my C100mk2 CAN actually be turned into Cinelog. It's a big process..

Canon Log -> 1D Lut Canon Log to Cineon -> 3D Lut, VFX Folder, Cineon to Linear -> 3D Lut, VFX Folder, Linear to BMDFilm -> BMDFilm to Cinelog-C

That said, the footage comes out weirdly exposed with overly-saturated greens and clipped reds when I convert to Rec. 709 or go from BMDFilm to Hunter's LUT.

Why?

[Andy600]

@dubzeebas - Everything up to the BMD Film to Cinelog-C lut is actually 1D and does not alter chromaticity (even though some of the Resolve luts you added are in the 3D VFX folder). This means your input colorspace is still that of Canon log (I presume CPLock Canon Log?) so when the BMD Film to Cinelog-C lut is added it is not transforming from BMD Film chromaticity (as it is designed to do) but something close to rec709 RGB. The transfer functions (log-to-log-to-linear-to-log1D parts that alter the tone curve) will be correct but you are mapping completely wrong RGB values to Cinelog-C RGB. BMD Film RGB is much wider gamut than Rec709 and this is why it will massively oversaturate and clip.

If I get a chance I will send you an advanced copy of the C100/C300 luts tomorrow - these will transform directly from Canon log (all the variants I mentioned in a previous post) to Cinelog-C in a single lut. The CP Lock version uses Canon's ACES IDT matrices.

[dubzeebass]

If I get a chance I will send you an advanced copy of the C100/C300 luts tomorrow - these will transform directly from Canon log (all the variants I mentioned in a previous post) to Cinelog-C in a single lut. The CP Lock version uses Canon's ACES IDT matrices.

Very interesting! Didn't know that was diff 1D & 3D was changing colorspace vs. just changing tone curve. Those advanced LUTS would be much appreciated!! PM'ing you my email address.

[Andy600]

@dubzeebass - sent

[dubzeebass]

Got it, thanks!

[2FAST]

Hi Andy,
I bought C100mk2 and color and contrast is different between original C100.

[Andy600]

@2FAST - you mean they produce different color even with the same in-camera settings for Canon Log? (specifically Canon Log/CP Lock)

The Canon Log luts we're releasing work with the C100, C300, 1DC and C500 - I'm curious to know what Canon has done with the C100 mark II if there is a noticeable difference. They haven't released an ACES IDT yet so I don't know the matrix coefficients to compare to the Mark I but AFAIK Canon log math hasn't changed - unless it's Canon log2 as in the C300?

[dubzeebass]

Andy check email, I sent a number of items.

[2FAST]

Hi Andy,

Yes color from original C100 between C100mk2 is different even it's same setting. Even Canon Log/CP Lock. Even wide DR. Both is C-log but not C-log2.
Primary color of mk2 has saturated more than mk1.
And Mk2 has cool(blue) and Mk1 has warmer color. Mk2 come video like. It's not only from WB shift I think.
I met man from Canon and asked why color was changed. He said engine is different.

I feel also log curve is bit different. But I'm really not sure and I didn't do any science test. But it's true that some mk2 user saying same.
I feel mk2 has a bit DR more in highlight but lost DR in shadow and middle. It's just I feel though. At least highlight roll off coming better.

Here is a compare movie.
https://www.youtube.com/watch?v=P6MNkV9d6FE

Someone says it's small difference but if I grading, I feel big difference.

But hey Andy, I was very busy in this half year and could back to this ML thread yesterday and surprised that Cinelog was progress so much! even I could not try yet. Finally now I'm downloading cinelog3 which was released in Jan.
So now you gonna make LUT for cinema eos? It's very nice to hear! Please let us know when it will be released!
Soon I will play with cinelog3 with 5D3 ML thanks for your work!

[janoschsimon]

big thx to andy for helping me to bring cinelog to my miniraw2 ;-) he is such a genius! UHHHH also nice for my 70d with magic lantern hehe

cheers janosch

[DeafEyeJedi]

Great news, @janoschsimon -- now curious to see your footage in slo-mo from your miniraw2! 

@Andy600 -- got a question for you... basically I'm hoping for a way to speed up my workflow within AE when using Cinelog-C:

https://vimeo.com/128048171

I remember you telling me that it is recommended to place the shadow/highlights fx plugin sandwiched in between the two OCIO in AE, correct?

Because I sometimes get better results if I were to instead place this on the bottom of the two OCIO (unless either I'm doing something wrong that somehow makes it look better or I'm literally blind and deaf) 

Also is there any way for me to have both the animation presets saved & enlisted without forcing me to open and install the configuration files each time I open AE?

Any suggestions or thoughts?

Much appreciated!

[janoschsimon]

@deafeyejedi

still some work to to :-)
ive messed up the linearsiation GRRRR but colors are thx to andys help SPOT on

YEAH

[lureb74]

Hi Andy600,

I'm trying to do the same steps you explained in this post (here's the link: http://www.magiclantern.fm/forum/index.php?topic=10151.msg140843#msg140843), but in AE.

I have some clips converted in Cinelog-C from 5D3-Cinelog-v3, then saved in Cineform444.

Now i'm trying in many ways to convert them using first OpenColorIO (conf: Cinelog-C, input: Cinelog-C, output: various) and then FilmConvert with many options, conformed to the option I previously choosed in OCIO: first with the 5D3 profiles, then the OCIO to sRGB and FConv to default, but I tried also the Arri profiles (as suggested), then the BMD4k, the RED Dragon, the Sony AS7, etc. Always selecting the corrisponding from OCIO...

The results are always superbs, but always different, in exposure, colors, curves, all is different! I expected the same results with all those conf. May be I was wrong?

Thank you for your attention

[Andy600]

@Lureb74 - I don't think Film Convert targets a unified colorspace before applying a film emulation so it will render differently depending on the source footage. Have you tried it with actual footage from the various cameras? for testing you can download some shots from CML's camera evaluations http://www.cinematography.net/.


Accurately matching color across different cameras is the intended function of ACES. ACES removes the aesthetic differences between each cameras color but for Magic Lantern Raw footage to work properly in an ACES environment we need different color matrices for each camera model and an ACES IDT - we need a Magic Lantern colorspace that provides a unified 'look' and that's what I am working on.

Magic Lantern currently uses color matrices from the Adobe DNG converter app (the DCRaw matrix is also from Adobe). The matrix describes to the raw reader (i.e. Resolve, Nuke, Premier etc) how to get from a cameras native RGB colorspace to XYZ. XYZ is used for connecting colorspaces and is part of DNG and CinemaDNG standards, plus it is used by .ICC profiles in Adobe apps.

The matrix used in MLV (gleaned from a set of 4 matrices but we only use one of them at present except for MLVFS and raw2cdng apps which have all 4) was derived from measuring sensor spectral sensitivity under a calibrated light source in a lab. The actual color was likely calibrated to an IT8 reference print target  - because a DSLR's primary function is stills photography.

IT8 targets are perfect for print and scanner calibration but we're dealing with images that are intended to be seen on a TV, monitor or Cinema screen. More importantly, we are, or should be, working to a standard (CinemaDNG) that is a little different and more restricted in terms of color metadata than DNG standard so a slightly different approach to color calibration is required (we can't use forward matrices or other 'tweaks' to get a good color fit because the extra metadata will just be ignored by most color apps) - I'll go into more detail when I publish the Magic Lantern Color Project.

Because Canon sensors have very good color metamarism and are linear where it counts, it is 'technically' possible to closely match a different cameras' default color (not it's dynamic range, noise levels etc) or, to a lesser extent, even a film stock but the initial aim will be to obtain a default color set that references a standard MacBeth colorchecker chart values in Rec709 colorspace using only 2 color matrices.

In test, we (Cinelog) have produced a much lower average mean error across 24 color patches using this method when compared to using the Adobe matrix and without any luts being needed. We also matched a 5D Mark III, 7D and a 50D with no detectable difference in the color patches on daylight shots in Rec709 colorspace (same lens).

[Danne]

Interesting read.
Are forward matrices used in ACR/after effects or will transcoding rely on color matrix 1 here as well?

[Andy600]

@Danne -The Forward Matrix tags describe the transform from white balanced XYZ to XYZ D50 (aka PCS) and are used in ACR - ACR also uses HSV tables for fine tuning color and correcting out-of-gamut color caused by the color matrix tags (the HSV table is embedded in a camera profile).

AFAIK, the Forward Matrix tags will be ignored by Premier and AE if the DNG is not debayered with ACR because they are (or should be) using CinemaDNG standard, not DNG standard - but it's Adobe so who knows? - it's easy to find out, just write one DNG with both the CM and FM tags and another with only the CM tags and see if there is a difference in color.

The color matrix tags are only intended for white balance but they also provide color if no FM tags are available. For MLV we really should be following CinemaDNG standard and have 2 color matrices (daylight and tungsten) calibrated for video, not print.

Forward matrices can be used to manipulate color saturation and hue but they are ignored in Resolve, Nuke etc.