How to output XYZ vor DCP?

Hi, I know, this topic is asked a lot, but I could not find an answer that really helps me here. And I have to find a solution for my client.

My Setting:
- I am grading a 90 min film with Epic and Canon DSLR footage for a client in DaVinci Resolve lite 9.
- I monitor on a JVC Broadcast monitor connected via a Decklink Extreme HD 3D.
- in Presets, "Monitoring format" is "HD 1080p25". "Video connection operates as" "YUV 4:2:2 SDI". Colorspace is "Video Levels".
- "Color Science" is set to "DaVinci YRGB"
- no LUTs are set (wether Input, nor Display, nor Output LUTs)

Now my client asked for XYZ Tiffs from which there will be made a DCP with EasyDCP Creator.

I did a render but the client said that in the theater the projected images had a severe greenish-brown tint.
The render was done with "Render to" set to Tiff. "Codec is" set to "Tiff XYZ 8 bit" and "unscaled full range data".

What do I have to do, so that the DCP that my client produces from my Tiffs will look like what I see on my broadcast monitor?

(Do I have to use Output and Display LUTs, or LUTs on a Track Node? And if yes, how?)

Please help. Kind regards,
Andreas

This is getting to be the $10,000 question.

Soon enough, of course, Resolve users will have the option to license an Easy DCP plugin, and hopefully a lot of those media translations will become part of the pre-set conversion process.

However, you are in the situation of needing to create a "DCDM" for use in creating a DCP package.

There are a couple of commercial packages available that can help you do this as a standalone solution -- either through Compressor, or straight out of a Final Cut timeline. Trying to manage the X'Y'Z' RGB conversion is probably not practical to attempt as it is really an exercise in reinventing the wheel, without any tools. You are already describing a less-than-ideal configuration, and for most intents, simply firing files blindly into space.

You have already discovered that a simple 8-bit XYZ TIFF doesn't hack it. There is an entire colorspace conversion (to P3) that must be negotiated, with a different gamma and white point, let alone that 8-bit is not nearly enough to survive large-format projection.

For the moment, do try to find something like a DCDM creator.

jPo

Hi JPO, thank you very much for your explanation.

Is it correct that since my broadcast monitor complies with Rec709 I am effectively grading in Rec709? Because Knut Eric Eversen writes in his blog entry ("Some ideas for best common practices when mastering DCPs")

A DCP could also be made from a file based master in 10 or 12 bit, Rec. 709 HDTV colorspace graded on a HDTV monitor. The Director of Photography should approve of how this looks in the movie theatre because a HDTV monitor is not ideal because you don’t typically have 14 footlambert and a dark room when grading on a HDTV monitor. But the results you can get from a 10 bit HD master is good enough for films like The Social Network so it could be great when done correctly.

How would I have to do the render in this case?
Would this work together with EasyDCP Creator?

If not, does anyone have a recommendation for a DCDM creator?
Could I somehow use Adobe After Effects for this?
(I am on Windows 7)

Kind regards,
Andreas

There is a 3D LUT in Resolve that does the 709 to XYZ colorspace and gamma transformation, you can apply it to the whole project via output LUT.

DCPCreator or openDCP and some other freeware does the transformation from 709 TIFFs - you can go ether way.

AE also comes with a XYZ colorspace output LUT - never tried that though.

There is a 3D LUT in Resolve that does the 709 to XYZ colorspace and gamma transformation, you can apply it to the whole project via output LUT.

DCPCreator or openDCP and some other freeware does the transformation from 709 TIFFs - you can go ether way.

AE also comes with a XYZ colorspace output LUT - never tried that though.

Without any calibration Frank is right: that will be your best bet to have a XYZ master that is sufficiently close.

However that will NOT guarantee that what you will see in projection is coherent with what you see in the monitor you have.

One day a group of DPs where in my room during a demo and one asked:
"why we have to pay so much money here for a DI where we can do it at [mom 'n pop shop] for half the price?"
"because I can guarantee that in the theater you will see what you signed of here..."
was my reply...
and that is the 10000$ somebody mentioned
We know how to calibrate and manage any color space and (as a company) can deliver exactly the right transformation to the final product. If anybody can do it I will go back to the farm I was born and go back to do butchering stuff... (not far that day I suppose...)

However, here some info.

When you talk about rec709 complaint, are you saying that that monitor is setup for BTU.1886 (vision environment for rec709 coded images) within a very narrow margin of tolerance/DeltaE? (and verifiable with an instrument?)

Or, (and that is a postmortem empirical way to close the circle), you can get an instrument, mapping the RGB/W/Gamma of your monitor and build a [my rec709]-to-[DCI_XYZ] conversion that will include all the deviance your viewing environment might have.

The Davinci transformation, I think is a standard rec709 to XYZ translation that keep as baseline the Idea that if you have a calibrated environment you will have a somewhat calibrated XYZ product viewed in a calibrated environment as well... ( I never use it because we have customs color space transformations/luts...)

Some pretty interesting thoughts Walter, thanks for sharing.

Here in my humble grading suite, getting halfway consistent color drives me nuts for a while.

On my side, I have a ton of variables in the game.

Premiere, After Effects and Resolve, all seem to handle color a bit different.
I feed the signal via SDI to BM HDLink that hands it over to a HP Dreamcolor.
I hardware calibrated the monitor, using HPs own hardware as good as possible.

Usually I set the monitor to 709 - but I found a ton of different descriptions on the web regarding the specs - brightness , gamma for 709. So I'm not sure, my signal chain is right at all.

On the other side, there are a ton of different display types from tablets to LED-TV and crappy client computer monitors. There is no way (that I know of) to give them all a similar experience. Just different gamma can ruin it completely. And then comes different software players on computers, that all show a different image.

Often I get a call from a client, that says the film I send them looks like sh..t - I tell him that it is well within the specs and looks great on my calibrated monitor. Then I hear, that EVERY blockbuster looks just fine when he watches it on his computer. Now the DCP thing opens a new can of worms.

How do you guys handle that?

I did a render but the client said that in the theater the projected images had a severe greenish-brown tint. The render was done with "Render to" set to Tiff. "Codec is" set to "Tiff XYZ 8 bit" and "unscaled full range data".

Sounds like a double XYZ conversion was done. EasyDCP Creator will do an RGB->XYZ conversion by default. If you supplied images already in the XYZ colorspace and the user didn't disable it in EasyDCP, the colors are going to look desaturated and have green/brown tints.

Also, it's not just converting from RGB->XYZ. You need a specific color profile, which takes into account the 2.8 gamma and DCI transfer function. I believe AE has DCI digital cinema profiles for XYZ.

I do have this kind of render when creating a DCP with files already in XYZ (conversion made under DaVinci). This is not desaturated, the green look is really thin but when we know how look the original image it's not the same.
Isn't it possible to let DaVinci make the conversion and then choose the bypass function under EasyDCP (to avoid a new conversion) ? I tested much things but never the same before/after look..!

I did a render but the client said that in the theater the projected images had a severe greenish-brown tint. The render was done with "Render to" set to Tiff. "Codec is" set to "Tiff XYZ 8 bit" and "unscaled full range data".

Sounds like a double XYZ conversion was done. EasyDCP Creator will do an RGB->XYZ conversion by default. If you supplied images already in the XYZ colorspace and the user didn't disable it in EasyDCP, the colors are going to look desaturated and have green/brown tints.

Also, it's not just converting from RGB->XYZ. You need a specific color profile, which takes into account the 2.8 gamma and DCI transfer function. I believe AE has DCI digital cinema profiles for XYZ.

2.8 gamma? which color space you use?

Any automatic conversion is implying a [starting color space] and an [ending color space].

Most of the "automatic RGB->XYZ" conversions use a rec709/BTU1886 primaries with display gamma (For BTu1886 is [0.64,0.33 - 0.3,0.6 - 0.15,0.06 - 0.3127,0.329 - gamma 2.4] ) for starting color space and XYZ primaries (1,0 - 0,1 - 0,0 - .3333 .3333 and 2.6 gamma parametric power function) for targeting color space.

If your starting color space is coherent with the BTU1886 standard it works.

All other cases will fail. Period.

Including if you are "close" to the standard but not quite there: it will still fail. More noticeable is the red shift I see in most plasmas HDTV: it is too orange (measured with a spectrophotometer and compared to the standard) and it will result in a DeltaE opposite when showed in the XYZ color space: will be too pink/magenta.

Some pretty interesting thoughts Walter, thanks for sharing.

Here in my humble grading suite, getting halfway consistent color drives me nuts for a while.

On my side, I have a ton of variables in the game.

Premiere, After Effects and Resolve, all seem to handle color a bit different.
I feed the signal via SDI to BM HDLink that hands it over to a HP Dreamcolor.
I hardware calibrated the monitor, using HPs own hardware as good as possible.

Usually I set the monitor to 709 - but I found a ton of different descriptions on the web regarding the specs - brightness , gamma for 709. So I'm not sure, my signal chain is right at all.

On the other side, there are a ton of different display types from tablets to LED-TV and crappy client computer monitors. There is no way (that I know of) to give them all a similar experience. Just different gamma can ruin it completely. And then comes different software players on computers, that all show a different image.

Often I get a call from a client, that says the film I send them looks like sh..t - I tell him that it is well within the specs and looks great on my calibrated monitor. Then I hear, that EVERY blockbuster looks just fine when he watches it on his computer. Now the DCP thing opens a new can of worms.

How do you guys handle that?

Frank, with your viewing environment you are probably already doing miracles...

Even with a perfect calibrated monitor (that is nearly impossible) there is a perception issue of viewing in a narrow field of view (about what, 6-10 degrees cone?) and a bright subject (the plasmas/LCD/Monitors often pump anything between 80 and 100 [cd/m2]) against probably some surrounding light where a DCP optimal viewing environment will cover more or less 45 degree cone view in a dark room with only 48[cd/m2] if you are in a DI suite.

This will translate in:
perceptual ( and measurable ) a LOT more contrast
perceptual a LOT more sharpness (due to the COC being 1/4 1/5 on the HD display)
perceptual ( and measurable ) less color overall due to the spread.

On top of it there are other viewing phenomenons: your iris will open more in a dim environment and that will make the sharpness even less, there there is the photometric adaptation to light that the eye/brain do that compensate for most of the problems and so on...

Then there is the best part...

The client DO NOT have photographic memory. They compare what they think they saw with what they see there.

I have in my room a plasma 50inc, a 12 meter screen that I can split in two and use a DI DLP grade 1 projection with the DCP XYZ material and in the other half a real film projector: all play together and sync and the match is within the technological limitation of the medium and the LUT used and still the client bit** about it. Even if you go to a store and you look at the same brand of HDtv the look different and it is the same bloody technology.

So, TL,TR, there are technological measurement and improvement that can be done to ensure that the final product is within strict tolerances, however, nothing can satisfy a stubborn client.

2.8 gamma? which color space you use?

Sorry, typo... should have said 2.6.

To the original poster, Andreas:

I think you and your client indeed did a double conversion. When using dcpcreator it's better to export full range rgb tiff and let easydcp do the conversion, cause it does a better job than Davinci.

That lut in resolve or the export option (tiff xyz space) is not working well translating the colorspace. Similar like the concersion in opendcp, which is also bad.

I had perfect results using Scratch to transform the colorspace and gamma, but that's off topic here.

Just export rgb tiff and let easydcp do what it's good at!

To the original poster, Andreas:


Just export rgb tiff and let easydcp do what it's good at!

That only work IF the original colorspace you use is the same easyDCP use for the transformation.... (And you're calibrated)

My Setting:
- I am grading a 90 min film with Epic and Canon DSLR footage for a client in DaVinci Resolve lite 9.
- I monitor on a JVC Broadcast monitor connected via a Decklink Extreme HD 3D.
- in Presets, "Monitoring format" is "HD 1080p25". "Video connection operates as" "YUV 4:2:2 SDI". Colorspace is "Video Levels".
- "Color Science" is set to "DaVinci YRGB"
- no LUTs are set (wether Input, nor Display, nor Output LUTs)

Seems like just the thing easydcp wants...

My Setting:
- I am grading a 90 min film with Epic and Canon DSLR footage for a client in DaVinci Resolve lite 9.
- I monitor on a JVC Broadcast monitor connected via a Decklink Extreme HD 3D.
- in Presets, "Monitoring format" is "HD 1080p25". "Video connection operates as" "YUV 4:2:2 SDI". Colorspace is "Video Levels".
- "Color Science" is set to "DaVinci YRGB"
- no LUTs are set (wether Input, nor Display, nor Output LUTs)

Seems like just the thing easydcp wants...

Not really: nobody is telling you that you are BTU1886 complaint/calibrated at all...

Hello,

After reading a 100 or so blog posts, I think I have this answered, but I have a few questions.

Background:

Shot Arri Alexa ProRes 4444 2k – ingest into Resolve – grade – output to Bluray DVD/web for festivals screeners and finally to DCP for digital projection at film festivals.

System config:

Resolve 10.1 – Decklink Extreme – HD-SDI via BNC cable – Panasonic BT-LH 2600W.

The process I should follow?:

I understand the difference between Rec709 video levels 64-940, and full range data levels 0-1023. From what I’ve read in the blogs, If I’m viewing through my HD broadcast monitor (the BTLH), then I need to set resolve to Rec 709 video levels 64-940). After grading, the DCP house says I should render out full range TIFFs (since they are 16 bit and DPX is lower, 12 bit, or 10 bit). The DCP house will then convert my blacks from 64 to 0 and my whites from 940 to 1023. The DCP house says they should able to match my look within a 4% margin of error.

My questions:

When I set Resolve to Rec709 FULL RANGE, instead of VIDEO LEVELS, I see the darker blacks, in my BTLH. What am I seeing? The image seems to represent more accurately what I’m seeing in my internal resolve scope. In other words, when I bring my blacks close to 0, and Resolve is set to rec709 full range, the image in my BTLH looks like the blacks are close to 0. When I set to video levels, my image looks washed out compared to what my internal scopes are representing. I know Resolve always works in 0-1023, but what am I seeing in the monitor? Can the BTLH represent 0-1023? I’m pretty sure not, but hoping some how it does.

Should I just work in Rec 709 video levels, use external scopes, and let DCP house adjust my blacks and whites? How will chroma be affected?

Thanks, you folks are a wealth of knowledge.

As you note, Resolve always works in 0-1023 internally. When you monitor video levels rather than data levels, it scales this to 64-940... but only for monitoring. When you go to render a full-range output, what you would see as 64 on external scopes when monitoring video levels should be at 0 in the resulting file. This means, unless I'm missing something, that it would be a mistake for the facility authoring the DCP to scale 64 to 0 and 940 to 1023 if you're giving them full-range TIFF files. That will result in shadow and highlight clipping.

However, it sounds like there's a potential issue with how your monitor is set up. What happens when you output color bars through Resolve? Do the PLUGE bars look correct?

Assuming you figure out the monitoring situation, you can cut a lot of uncertainly out of the DCP authoring process by just delivering full-range XYZ TIFF files (using the 'Rec. 709 to DCI-XYZ' LUT included with Resolve 10) and telling the facility authoring the DCP not to do any gamma/color conversion at all.

hello,
Juste a few word to warn you about this rec709 to DCI XYZ LUT.
From several test I've made with different footage it would always result in a greenish look, while the new DCI jpeg2000 export option would perform a correct XYZ conversion(yet the Jpeg2000 standard from davinci doesn't seem to be DCI...)
(tested with neodcp player)

I'm not very confortable with all toose lut and colorspace thing, so the greenish look may be due to a mistake on my part, yet I'd use it carefully if not acquainted.

LightSpace will generate a LUT that is mathematically correct is you want to do one yourself...

http://www.lightillusion.com/lightspace.html

davinci resolve XYZ transform made via the dci jpeg2000 export seems good enough.
I just hope they will make a jpeg2000 profile 3 or 4 in next release. This would mean exporting to open DCP directly from davinci.

edit
here is a little test I just made :
thoose are two mixed screen captures, DCI Jpeg2000 with no lut applied on the left, and tiff with rec709 to xyz lut applied. The footage is from a red MX played with stereoscopic player.

So is the current JPEG 2000 Export Option of Resolve Resolve 11 DCI Compliant? My Jpeg Sequences played well in EasyDCP Player...

So is the current JPEG 2000 Export Option of Resolve Resolve 11 DCI Compliant? My Jpeg Sequences played well in EasyDCP Player...

Can anyone confirm if Resolve 11.2 exports DCI compliant JPEG 2000?

Can anyone confirm if Resolve 11.2 exports DCI compliant JPEG 2000?

Hi tobytomkins,
from my last test, jpeg2000 encoded via davinci resolve was still profile 1 while DCI need profile 3 or 4.
if you make a dcp via a resolve jpeg2000 sequences easyDCP player will prompt an error about that profile problem.

see this page :
http://www.digitalpreservation.gov/form ... 0138.shtml

[edit] Didn't read your message carefully, haven't tested yet on 11.2

[Peter Chamberlain]

Try using the easyDCP plugin to make your DCI compliant DCPs.