Very big trouble. DCP

[Avi Liani]

DPX from Color I'd thing, if you are in a Final Cut workflow.

I can't remember if Color or Resolve output OpenEXR, but I like 16 bit OpenEXR more than DPX if you can get it. That is how I output a lot of composites. I don't think DCP can use OpenEXR as input though.

TIFF is great too just make sure you are outputting real 16bit TIFF files. I thought Quicktime did 16bit ... but I haven't used Quicktime for image file export in so long ... I just don't know anymore.

Shake, After Effects and Nuke - compositing applications in general, are good for image sequence output.

thanks and what about FCP is it good to export image sequence.

and How can I export image sequence out of Color. Because color does not recognize video filters and transitions, texts added in FCP

And how can I check if image is 16 bit or less and why bits are so important?

 

[Alexander Ibrahim]

thanks and what about FCP is it good to export image sequence.

and How can I export image sequence out of Color. Because color does not recognize video filters and transitions, texts added in FCP

And how can I check if image is 16 bit or less and why bits are so important?

OK, so what I am hearing is that you've finished in Final Cut already. Is that right? If so you have three choices:

1) Export TIFF files directly from Final Cut Pro using Quicktime conversion.
You have to use a custom setting, and export 24fps with "Millions+" color.

I have never evaluated Final Cut Pro's image sequence export with high quality data sources.

2) Export ProRes444 and use another application, like Motion or After Effects to convert that to DPX or some other format acceptable for DCP.

ProRes4444 (or Cineform FilmScan 444) is the highest quality footage you can get out of Final Cut Pro, its just 12 bit.

You add another transcoding step to your output this way, which is generally undesirable.

3) Take your project into another finishing system and recreate your titles and transitions.

This is probably the best route. You can conform your footage to R3D originals, get a new color correction and output up to 4K DPX or TIFF from which to create your DCP

I recommend Resolve for this on the budget side. Incredibly powerful tools, and for a project that is "mostly" finished you'll be more than satisfied with the editorial controls.

Smoke is really my tool of choice though for finishing ... not that I have ready access to Smoke. Primarily because the compositing and editorial tools are so much better. (That's usability better not technical better- sometimes they are technically better, sometimes they aren't. That discussion is way beyond the scope of this post.)

Anyway, while I am sure this will give the best results, it will also be the most work. I suppose its the lesson of the Rubik's Cube - sometimes you have to go backwards a bit before you can go forward.

Finally, about the bits! How to give a compressed lesson in computer science and color.

Well more colors are good right? More bits means that the footage can represent more shades of color. Each bit you add multiplies the whole thing by two.

So, 8 bit gives you 256 shades of the primary colors: red, green and blue.

9 bit doesn't exist- for arcane computery reasons we only use even numbers of bits for things. If it did exist, it would allow you to represent 512 shades of each primary color.

10 bit gives you 1024 shades of the primaries.

12 bit gives 4096 shades of the primaries.

Again for arcane computery reasons we don't use 14 bits. (Its helpful to remember that most things are broken into 8 bit chunks, so computationally 16 bit is just as easy as 14 bit for the computer to handle.)

16 bit gives 65536 shades of the primaries. That's 281 trillion colors ... and its plenty for any approximation of human vision.

To explain that last bit of apparent overkill, humans only see between a few million and for the very gifted (tetrachromats) a few billion colors. The thing is that these colors aren't evenply spread out when you graph their values. Because of that the computer needs LOTS more ability to discriminate color to actually encompass everything humans can see.

You only need 12-bit color for DCP at present, which is fine because that's all Red One can give you.

Epic will use a 16 bit "container" but even it will only deliver about 14 bits of color - because of limitations in sensor engineering. (signal to noise ratios). Because of something called perceptual coding, it will be a very good match for the colors we can see, the best motion picture camera ever in fact, yet still not able to match perfectly.

 

[Avi Liani]

OK, so what I am hearing is that you've finished in Final Cut already. Is that right? If so you have three choices:

1) Export TIFF files directly from Final Cut Pro using Quicktime conversion.
You have to use a custom setting, and export 24fps with "Millions+" color.

I have never evaluated Final Cut Pro's image sequence export with high quality data sources.

2) Export ProRes444 and use another application, like Motion or After Effects to convert that to DPX or some other format acceptable for DCP.

ProRes4444 (or Cineform FilmScan 444) is the highest quality footage you can get out of Final Cut Pro, its just 12 bit.

You add another transcoding step to your output this way, which is generally undesirable.

3) Take your project into another finishing system and recreate your titles and transitions.

This is probably the best route. You can conform your footage to R3D originals, get a new color correction and output up to 4K DPX or TIFF from which to create your DCP

I recommend Resolve for this on the budget side. Incredibly powerful tools, and for a project that is "mostly" finished you'll be more than satisfied with the editorial controls.

Smoke is really my tool of choice though for finishing ... not that I have ready access to Smoke. Primarily because the compositing and editorial tools are so much better. (That's usability better not technical better- sometimes they are technically better, sometimes they aren't. That discussion is way beyond the scope of this post.)

Anyway, while I am sure this will give the best results, it will also be the most work. I suppose its the lesson of the Rubik's Cube - sometimes you have to go backwards a bit before you can go forward.

Finally, about the bits! How to give a compressed lesson in computer science and color.

Well more colors are good right? More bits means that the footage can represent more shades of color. Each bit you add multiplies the whole thing by two.

So, 8 bit gives you 256 shades of the primary colors: red, green and blue.

9 bit doesn't exist- for arcane computery reasons we only use even numbers of bits for things. If it did exist, it would allow you to represent 512 shades of each primary color.

10 bit gives you 1024 shades of the primaries.

12 bit gives 4096 shades of the primaries.

Again for arcane computery reasons we don't use 14 bits. (Its helpful to remember that most things are broken into 8 bit chunks, so computationally 16 bit is just as easy as 14 bit for the computer to handle.)

16 bit gives 65536 shades of the primaries. That's 281 trillion colors ... and its plenty for any approximation of human vision.

To explain that last bit of apparent overkill, humans only see between a few million and for the very gifted (tetrachromats) a few billion colors. The thing is that these colors aren't evenply spread out when you graph their values. Because of that the computer needs LOTS more ability to discriminate color to actually encompass everything humans can see.

You only need 12-bit color for DCP at present, which is fine because that's all Red One can give you.

Epic will use a 16 bit "container" but even it will only deliver about 14 bits of color - because of limitations in sensor engineering. (signal to noise ratios). Because of something called perceptual coding, it will be a very good match for the colors we can see, the best motion picture camera ever in fact, yet still not able to match perfectly.

Thanks for comprehensive explanation

I am right now exporting 16 bits from compressor, So I will make new DCP.
So DCP will automatically reduce to 12bit?

 

[Alexander Ibrahim]

Thanks for comprehensive explanation

I am right now exporting 16 bits from compressor, So I will make new DCP.
So DCP will automatically reduce to 12bit?

I think that DCP supports 12, 14 and 16 bit formats as inputs, but it only uses 12 bit as output. That may be DCP application dependent.

DCP and camera formats share the notion of an "optimized palette."

So, its not just "throwing away" the extra data, its using it to find the best match to the DCI-P3 colorspace and encoding that in the 12 bit files it creates.

In other words, its still useful.

 

[Dan Hudgins]

65535 and a folder

65535 and a folder

Another trick you can use on OS that are limited to 65535 about files in a folder is to copy the first made files out of that folder and put them in another folder while new ones are being added, on a multi-tasking OS you can do that, just run the program making the frames in background and use the OS file manager to copy the first made files to another folder (cut and paste, not just copy since the idea is to free filenames in the folder being used by the program that is making the frames).

I have something called "Divided" filename in my software that only puts 500 files in a folder and automatically makes new folders, so you get faster access and don't run into the ~65535 files per folder limit. On some OS more than ~511 files in folder slows down the I/O quite a bit, so by my keeping at 500 per sub-folder I can run hours of frames under even DOS...

 

[M Most]

I think that DCP supports 12, 14 and 16 bit formats as inputs, but it only uses 12 bit as output. That may be DCP application dependent.

DCP and camera formats share the notion of an "optimized palette."

So, its not just "throwing away" the extra data, its using it to find the best match to the DCI-P3 colorspace and encoding that in the 12 bit files it creates.

In other words, its still useful.

There is no "input" format when making a DCP, there is only the format you're creating. Different DCP authoring softwares support different input formats by providing specific color matrices and conversion parameters, but the DCP format itself is a distribution format that is described as 12 bit. What you use to create it has nothing to do with the DCP format description.