Knut Erik Evensen – KESE

Sometimes you may want to add something to a DCP trailer or feature film.
Example: Trailers often need new graphics in a new language.

In Photoshop and Resolve you can apply 3D LUTS to an image to match the white point of the original DCP.

If the DCP has a warm white you could use a D60 3D LUT and if it has a cooler white you can use a D65 3D LUT.

If you render a video file in Final Cut X or Premiere with graphics that has a X’Y’Z’ 3D LUT applied, remember to choose no color transform in Easydcp and similar DCP mastering software.

If you apply a rec709 to DCI X´Y´Z´ 3D LUT to a video or image in Resolve that is used on a DCP, change Project settings – Color Management – Timeline Colorspace  to DCI X´Y´Z´ .

You can apply 3DL 3D LUTs in Easydcp Creator.

You can apply Cube 3D LUTs in Resolve and Photoshop.

How to apply a 3D LUT in Photoshop
Layer – New adjustment Layer – Color lookup

 

Click on load 3d lut and choose a 3D LUT. You can now use a cube 3D LUT.

 

Some 3D LUTS

Here are some P3 and REC709 3D LUTS in 3DL and CUBE (float) format using the Nuke 3D LUT method from this post and the matlab NPMs from this post.

P3 D61 2.6 gamma to X’Y’Z’ 7z

White in the 12 bit 3dl file is 3885 3960 3997 It is the correct D61 value in SMPTE EG-432-1-10

 

P3 D60 2.6 gamma to X’Y’Z’ 7z

White in the 12 bit 3dl file is 3886 3960 3972. It is the correct D60 value in SMPTE EG-432-1-10

 

P3D65 2.6 gamma to X’Y’Z’ 7z

White in the 12 bit 3dl file is 3883 3960 4092
It is the correct D65 value in SMPTE EG-432-1-10

 

P3 DCI WHITE 2.6 gamma to X’Y’Z’  7z

White in the 12 bit 3dl file is 3794 3960 3890 (greenish)
It is the correct DCI WHITE value in SMPTE EG-432-1-10

 

REC709 D65 2.2 2.4 2.6 gamma to X’Y’Z’ 7z

White in the 12 bit 3dl files is 3883 3960 4092
It is the correct D65 value in SMPTE EG-432-1-10

 

REC 709 D60 2.2 2.4 2.6 gamma to X’Y’Z’ 7z

White in the 12 bit 3dl files is 3886 3960 3972. It is the correct D60 value in SMPTE EG-432-1-10

 

How to calculate RP-177 3×3 matrices in Matlab.

Update: For part 2 with the RP-177 TRA matlab code see this post

The Matlab code is made by Tinna Lif Gunnarsdottir.

To make DCPs in DCI-P3 D6O, D61, D65 or Rec 709 D60, D61, D65 you can use SMPTE RP-177:1993 3×3 NPMs to transfer from RGB to XYZ.

NPM = normalized primary matrix

For more info on the math see SMPTE EG 432-1-10 and How to make 3D LUTs in Nuke (R´B´G´to X´Y´Z´)

update:

Here are some 3D LUTS made with these 3×3 matrices

 

Example: DCI-P3 D6O

Here is the matlab code to calculate the correct RP-177 RGB-XYZ matrices to go from DCI-P3 D6O to XYZ.

MATLAB CODE
#DCI-P3 Primaries

R = [0.68 ; 0.320 ; 0];
G = [0.265 ; 0.690 ; 0.045];
B = [0.150 ; 0.060 ; 0.79];

P = [R G B];

#DCI- P3 D60 Whitepoint

X = 0.3217;
Y = 0.3378;
Z = 0.3405;

W = [X/Y ; 1 ; Z/Y];

CC = P \ W;

C = [ CC(1,1) 0 0 ; 0 CC(2,1) 0 ; 0 0 CC(3,1)];

NPM = P * C

NPM =

0.504739646629896 0.264744009223770   0.182855006076470 0.237524539590539 0.689333457978873   0.073142002430588
0                                               0.044956529868187   0.963036365336076

END MATLAB CODE

Pasted into a Nuke color matrix node:
0.50473965   0.26474401     0.18285501
0.23752454    0.68933346    0.073142
0                            0.04495653   0.96303637

The middle row sum to 1
0.23752454+0.68933346+0.073142=1

Other 3X3 NPM matrices

The Rec709 and DCI-P3 D65 3×3 matrices is also used in this post

DCI-P3 D65

WHITE POINT
X = 0.3127;
Y = 0.3290;
Z = 0.3583;

NPM from matlab:
0.486570948648216 0.265667693169093 0.198217285234362
0.228974564069749 0.691738521836506 0.0792869140937450
0                                               0.045113381858903 1.043944368900976

Pasted into a Nuke color matrix node:
0.48657095  0.26566769  0.19821729
0.22897457   0.69173852    0.07928691
0                           0.04511338    1.04394437

The middle row sums to 1
0.22897457+0.69173852+0.07928691 = 1

DCI-P3 D61

WHITE POINT
X = 0.3198;
Y = 0.3360;
Z = 0.3442;

NPM from matlab:
0.500852913533834 0.264968867481203  0.185963933270677
0.235695488721804 0.689918937969925  0.0743855733082713
0                                              0.044994713345865  0.979410048558897

Pasted into a Nuke color matrix node:
0.50085291 0.26496887 0.18596393
0.23569549 0.68991894 0.07438557
0                          0.04499471 0.97941005

Middle row sums to 1
0.23569549+0.68991894+0.07438557=1

DCI-P3 DCI-WHITE

WHITE POINT
X = 0.314;
Y = 0.351;
Z = 0.335;

NPM from matlab
0.445169815564552 0.277134409206778   0.172282669815565
0.209491677912731   0.721595254161044    0.068913067926226
0                                              0.047060560053981 0.907355394361973

Pasted into a Nuke color matrix node:
0.44516982 0.27713441     0.17228267
0.20949168 0.72159525    0.06891307
0                         0.04706056   0.90735539

Middle row sums to 1
0.20949168+0.72159525+0.06891307=1

REC 709
REC 709 D65, D60, D61 fits inside DCI-P3 and can be used in DCPs.

REC709 Primaries
R = [0.640 ; 0.330 ; 0.030];
G = [0.300 ; 0.600 ; 0.100];
B = [0.150 ; 0.060 ; 0.790];

REC 709 D65

D65 WHITE POINT
X = 0.3127;
Y = 0.3290;
Z = 0.3583;

NPM from matlab:
0.412390799265959 0.357584339383878 0.180480788401834
0.212639005871510   0.715168678767756   0.072192315360734
0.019330818715592   0.119194779794626   0.950532152249661

Pasted into a Nuke color matrix node:
0.4123908  0.3575843   0.18048079
0.21263901 0.71516867 0.07219232
0.01933082 0.1191948   0.95053215

The middle row should sum to 1 according to RP-177 so I changed 0.71516868 to 0.71516867.

REC 709 D60

D60 WHITE POINT
X = 0.3217;
Y = 0.3378;
Z = 0.3405;

NPM from matlab:
0.431575918832227 0.355727231008106 0.165035512089804
0.222531333147867 0.711454462016212   0.066014204835922
0.020230121195261 0.118575743669369  0.869187030339634

Pasted into a Nuke color matrix node:
0.43157592 0.35572723 0.16503551
0.22253133 0.71145447  0.0660142
0.02023012 0.11857574 0.86918703

Changed 0.71145446 to 0.71145447 so the middle row sums to 1
0.22253133+0.71145447+0.0660142= 1

REC 709 D61

D61 WHITE POINT

X = 0.3198;
Y = 0.3360;
Z = 0.3442;

NPM from matlab:
0.427462229871868 0.356162746222987 0.168160738190859
0.220410212277682  0.712325492445975 0.067264295276343
0.020037292025244 0.118720915407662 0.885646554471856

Pasted into a Nuke color matrix node:
0.42746223 0.35616275 0.16816074
0.22041021  0.71232549 0.0672643
0.02003729 0.11872092 0.88564655
Middle row sums to 1
0.22041021+0.71232549+0.0672643=1

 

Add silence
Sometimes audio files are too short and does not match the length of the video files
Instead of making another reel with a silent soundtrack you can add some silence to the audio files using Sox.

To make a 48 khz 24 bit 10 second silent wav file with sox:
sox -n -r 48000 -b24 silent10sec.wav trim 0.0 10.0

To add this to the end of another 48 khz 24 bit sound file with sox:
sox film.LFE.wav silent10sec.wav film2.LFE.wav

 

Extract Channels

Extract channels from interleaved 5.1 sound files
If the file is L, R, C, LFE, LS, RS
sox film51.wav left.wav remix 1
sox film51.wav right.wav remix 2
sox film51.wav center.wav remix 3
sox film51.wav lfe.wav remix 4
sox film51.wav ls.wav remix 5
sox film51.wav rs.wav remix 6

Installation
In Ubuntu Linux you install sox using sudo apt install sox

On Windows you can install sox using Bash on Ubuntu on Windows and install sox with apt install sox or get the binaries from the Sox homepage.

On a Mac you can install sox using homebrew and install sox with brew install sox or get the binaries from the Sox homepage.

Here are some tips on downloading and uploading DCPs and other large files like prores quicktime files:

Downloading
-I recommended using a fast wired internet connection. (20/20-50/50 mbit/s). (Use a good network cable/router)
-I recommend using a download manager like Free download manager  (Windows/mac) that support resume and multi-part downloading. Free Download will allocate disk space in the start of the transfer and then the download speed will increase.

Uploading
-I recommended using a fast wired internet connection. (20/20-50/50 mbit/s). (Use a good network cable/router)
-I recommend a reboot of the computer or at least a restart of chrome before uploading large files (100 GB) in a web uploader.

Filemail

I have used filemail to receive files on many film festivals I have worked on.

-On filemail.com you can send 30 GB files for free. With a Filemail Pro or Business account you can send and receive 100-200 GB files.

-I have used the filemail html5 uploader on many 100 gb files. You can also use the html5 uploader to send folders. The Desktop filemail app also works well on 100 gb files and can be used to auto download all incoming files.

-With a Filemail business account you will have access to the files for longer and people can use a branded upload web page. You can also enable the md5 hash value.

Filemail md5 hash check

Filemail shows the md5 checksum on the download page if you enable it in the settings for the business download page.

 

This way you can check if the uploaded file is correct if the client sends you the original md5 checksum. You can ask for the md5 checksum as info on the upload page. This way you and the client can verify that the file is correct.

Generating a MD5 checksum
The client can find the MD5 checksum of a file with utilities like Hash Tab for windows and Hash Tab for mac

You also use the command md5 "file" in a terminal on a Mac to generate a md5 file hash value. The Terminal app is in Applications – Utilities. You can type md5 and push space, then drag the file from a finder window and push enter.

You can use the command certutil -hashfile "file" md5 on a Windows 7, 8, 10 computer in a command prompt (CMD)

You can use md5sum "file" in a terminal on a linux computer.

Verification
-If the sender sends a 7z archive, the receiver will know when unpacking if the file fails the 7z checksum test.
-When sending prores quicktime files it is important that the sender include a md5 checksum to enable verification of the file.
-You can combine these two Verifications methods.
-DCPs have the checksums in the XML files that can be verified with easydcp player or using this command
openssl sha1 -binary "FILE_NAME" | openssl base64 

7z archives
-If you want to send a uncompressed quicktime file or wav files I suggest using 7z with normal compression. A uncompressed quicktime file will be reduced to half the size after 7z compression.
-I recommend Keka on mac or 7zip on Windows.
-7z archives can be protected with encryption and a password and have a built in hash check.
-In Keka, choose 7z and store for no compression or normal for normal compression, then drag the folder to the Keka window.
-You can also split the archive in parts and upload the file in parts. Example 20 GB. If one of the parts is corrupted you only need to upload that part again.

Use a PAR2 recovery file as backup
-You can use a PAR 2 file as an additional backup solution if there is no time to upload/download the corrupt files again . I have tried both MultiPar and MacPAR Deluxe to generate PAR 2 1 percent redundancy files. PAR 2 files with 1 percent redundancy are 1 percent of the size of original file. If the downloaded file is 0,5 percent corrupted, you can use the 1 percent PAR 2 recovery file to recover the file instead of downloading again. I have tried this and it works.

Amazon S3

Amazon S3 can be used to deliver and receive big files on the internet.
-You can use the s3 web uploader or use CrossFtp and similar programs to upload to S3. Crossftp Pro can use multipart and integrity checks when uploading so you will know the uploaded file is working.
-After uploading you can make the link accessible (public) and give the clients the S3 download link.
-S3 storage is very reliable and scalable. You pay for each download.
-If you want a client to upload a file to a s3 upload bucket you can make an upload user in https://console.aws.amazon.com/iam/home#users with a policy that enables the user to upload to that upload bucket.

Like this policy:

From http://docs.aws.amazon.com/AmazonS3/latest/dev/example-policies-s3.html#iam-policy-ex0:
{
"Version": "2012-10-17",
"Statement": [
{
"Sid": "VisualEditor0",
"Effect": "Allow",
"Action": [
"s3:ListBucket",
"s3:GetBucketLocation"
],
"Resource": "arn:aws:s3:::examplebucket"
},
{
"Sid": "VisualEditor1",
"Effect": "Allow",
"Action": [
"s3:PutObject",
"s3:GetObject",
"s3:DeleteObject"
],
"Resource": "arn:aws:s3:::examplebucket/*"
},
{
"Sid": "VisualEditor2",
"Effect": "Allow",
"Action": "s3:ListAllMyBuckets",
"Resource": "*"
}
] }

– then give the client the access key and secret key for the upload user. (with this policy the user will see all buckets, but can only upload to the examplebucket bucket.

 

Aspera Cloud
I have used Aspera Cloud on many DCP transfers. It is used by many film festivals and post-production services. It is very fast.

Google Drive
i have downloaded many DCP with free download manager from Google Drive. It is fast.

Dropbox
i have downloaded many DCP with free download manager from Dropbox. I had no problems.

DCP and quicktime specific advice

-When downloading a DCP folder in Filezilla from a FTP server, I downloaded the XML files and the rest of the files as binary to get them to pass the file hash check in Easydcp Player.

-Ask people who submit DCPs to use the archive format 7z instead of zip. This way you can avoid the (large) Mac OS archive utility zip files that can only be opened by the Mac OS archive zip utility. Both Keka and 7-Zip can make 7z archives.

-Sometimes I get a error message when opening a quicktime file on windows pc that was made on a mac. If I rename the file it works.

-I recommend checking DCPs in Easydcp player. The trial version will let you know if the DCP fails some checks. You can also run a file hash check.

You should also make a DCP quality checklist that may include:
-Is the DCP named correctly?
Is it in the correct scope or flat format? The DCP should have a correct digital naming convention name. Example: The DCP is called Dcpname_SHR_S but is actually Dcpname_SHR_F because the subtitles are burned in too low and will be cropped at a scope preset in your cinema.
-Is the DCP in the correct language/subtitle language?
-Will the DCP play at your venue. Example: You may have a server with old firmware that can play 25 SMPTE DCPs with burned in subtitles, but not SMPTE 2010 subtitles. (all film festivals have updated firmware).

-If you have downloaded a DCP folder archive like DCP.7z on a mac and want to ingest in on a digital cinema media block/server/player you can extract it to a DX115 hard drive or similar.

Other tips

-If you get an error message extracting a zip file in Keka that says error code 2 using p7zip, you probably need to extract the zip file with mac os archive utility.

12 bit monitoring is great when grading LOG for digital cinema (DCP) on a 12 bit digital cinema projector.

To have the correct black level and correct contrast ratio you will need the correct review room contrast ratio in the cinema: 1500:1, 100:1 intra frame. Movie theaters have 1200:1.

You will also need the correct white light level (14 fL) in the cinema to get the contrast correct.

You will also need a dark room.

See this post for more info on good projection.

Resolve 11 beta 2/Desktop Video 10.1.3 adds dual link 3GB HD-SDI DCI 12 bit monitoring with the Decklink 4K Extreme and the BlackMagic Design UltraStudio 4K

12-bit 4:4:4 video monitoring support with Decklink 4K Extreme and Ultrastudio 4K (up to Stereo 2K DCI)

-What’s new in Resolve 11 Beta 2
BlackMagic resolve website

In this post there are more 3D LUTS

Updated in 2016 with some clarifications

The built in SRGB TO XYZ conversion used in some DCP conversion software do not have the correct color science.

See this post for more info.

The included XYZ 3D LUTs in this post are proof that you can generate 3D LUTs using the color science math from the REC709 example on page 42 in the “SMPTE Digital Cinema White Gamut Practices Study Group Report” by Rod Bogart and Rick Sayre.

The R´G´B´ values 4095, 4095, 4095 ends up as 3883 3960 4092 in X´Y´Z´

 

 

The “SMPTE Digital Cinema White Gamut Practices Study Group Report” (Link) from 2009 concluded that the color science used to make some commercial DCPs around 2008 was not correct (out of gamut).

The color science is explained more in SMPTE EG 432-1:2010

They explain how to use D55, D60, D61 and D65 as mastering whites and how these white points fit inside X´Y´Z´. They explain that DCI White is not necessary the mastering white.

With SMPTE 431.2:2011 they made a new projector configuration file (PCF) standard that is called Virtual White.

The DCI specifications  says that the projectors in cinemas should follow SMPTE 431.2:2011.

 

Update 2016

The color science math do get to X´Y´Z´ in this post produces the correct values for white in X´Y´Z.

In NUKE you can make 3D LUTs for Resolve and other software that have 3×3 SMPTE RP-177 color matrix conversions, gamma conversions and multiplications baked in.

The generateLUT node is now disabled in NUKE non-commercial (formerly PKL).

Links:
Nuke

http://help.thefoundry.co.uk/nuke/content/reference_guide/color_nodes/generatelut.html

Blackmagic Davinci Resolve

Example 1:
Making a REC709/SRGB R´G´B´to DCDM X´Y´Z´ 3D LUT for Resolve:

How to make a R`G`B` to X´Y´Z´ 3D LUT in NUKE:

 

-In the CMSTestPattern node I used a Cube size of 65, alternatives are 33 or 17.

-Linearize the R´G´B´ to RGB with a gamma function in a Gamma node: Example: 2.4 simplified gamma function= 0.416667
Alternative: SRGB 2.2 simplified gamma function=0.45

-Apply the rec709/SRGB RGB XYZ to RP-177 3×3 NPM color matrix in a ColorMatrix node.

The NPM from matlab after doing the RP-177 math is:
0.412390799265959 0.357584339383878 0.180480788401834
0.212639005871510   0.715168678767756   0.072192315360734
0.019330818715592   0.119194779794626  0.950532152249661

The middle row sums to 1.

The NPM after copying and pasting into the Nuke colormatrix node:

0.4123908  0.35758434   0.18048079
0.21263901 0.71516867   0.07219232
0.01933082 0.11919478   0.95053215

This is very similar to the example.

The middle row should sum to 1 according to RP-177 so I changed 0.71516868 to 0.71516867.

-Apply the normalizing constant multiplication in a multiply node. Luminance (48) divided by the normalizing constant (52.37) = 0.91655527
After pasting it into nuke it is shorten to:
0.916555

-Apply the DCP 2.6 gamma in a Gamma node to get X´Y´Z´ values.

-In the GenerateLut node I put: .3dl, 12 bit in/out, 3dl file format
autodesk flame/smoke.
Alternative: 16 bit.

The values for 4095,4095,4095 in the resulting 3D LUT (12 bit 3dl integer file) is:
3883 3960 4092

This is the correct result for D65 in X´Y´Z´
and it is the same result as in the example conversion.

 

With this 3D LUT you are inside the Virtual white X´Y´Z´ gamut (DCI/SMPTE 431.2:2011 compliant).

To convert this 3D LUT to a floating point .Cube 3d LUT that is compatible with Photoshop, Resolve and After Effects I used opencolorio.
In terminal on a mac I use this command to convert the 3D LUT:

ociobakelut -lut srgbtoxyz.3dl -format iridas_itx -cubesize 65 srgbtoxyz.cube

to install opencolorio/ociobakelut I used these commands on a mac:

/usr/bin/ruby -e “$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)”

brew doctor

brew install opencolorio

Download a 7z archive with the REC709/SRGB 3D LUTS files
It includes  2.2, 2.4, 2.6 CUBE and 3DL 3D LUTS (The cube 3d luts are in float)

Usage in Resolve (updated for Resolve 11)

The CUBE LUT can be placed in a folder you call Nuke or DCP inside Resolve´s LUT folder:
On a Mac: /Library/Application Support/Blackmagic Design/Davinci Resolve/LUT
On a PC: C:\ProgramData\Blackmagic Design\DaVinci Resolve\Support\LUT\

In Resolve you can apply it to the timeline from:
Project settings – Timeline Lookup tables – 3D Output Lookup table
or you can use a corrector node

You can render DCPs directly from Resolve with the Resolve Easydcp plugin. If you choose DCI X´Y´Z´ in the timeline color space no color transform will be applied and you can use the 3D luts to DCI X´Y´Z´ from this post.
See this post for more info.

Or export to another format and use it in Fraunhofer Easydcp .

Usage in After Effects
The resulting 3D LUT converted to .CUBE can be used in a non-linear 32 bit Adobe After Effects CC project
with the apply color LUT effect
when exporting 16 BIT TIFF or similar
or DCI jpeg2000 sequences with the Fnord DCI JPEG2000 encoder.

Usage in Easydcp
The 3DL 3D LUT can be applied in Easydcp.

In Easydcp Creator 2.2 you have the option to make custom color transformations. You can modify the 3×3 matrix, gamma, normalization constant/multiplication to suit your workflow. This is better than using 3D LUTs that has these color transformations baked in.
Read more about the custom color transformations in this post.

Example 2.

P3D65 to X´Y´Z´
The same example as I used in this post.
This 3D LUT can be used with Arri Alexa material that has the
C-log to P3 D65 3D LUT applied.
Link:Arri lut generator

 

Changed the color matrix node to:

The DCI-P3 D65 NPM from matlab after doing the RP-177 math:

0.486570948648216 0.265667693169093 0.198217285234362
0.228974564069749 0.691738521836506  0.0792869140937450
0                                               0.045113381858903   1.043944368900976

The NPM after copying and pasting into the Nuke colormatrix node:

0.48657095  0.26566769  0.19821729
0.22897457  0.69173852    0.07928691
0                             0.04511338    1.04394437

The middle row sums to 1
0.22897457+0.69173852+0.07928691 = 1

It is similar to the P3 D65 NPM from SMPTE EG 432-1-10

0.48657;0.26567;0.19822
0.22897;0.69174;0.07929
0.00000;0.04511;1.04394

Changed the gamma function node to 1/2,6=0.384615385 for P3
It becomes 0.384615 after pasting into Nuke.

Example P3 D65 to X´Y´Z´3D LUTS:

Download 7z archive with the P3 D65 3D LUTS files
It includes CUBE AND 3DL 3D LUTS (The cube 3DLUTS are in float):