CC problematic and filmout

[Rainer Fritz]

Hi !

We are in the final step with two features and are grading them at the moment one on nucoda filmmaster and the other on lustre.

First thing that we see is, that the logarithmic output of redline is not compatible with any logarithmic workflow set up for filmscans and LUTs. Is there any chance to get a real log dpx out of RED ? Seems like we are forced to do a linear grade and log conversion after grading before or at filmout..... first try was redspace/redlog, redspace/pdlog685 looks also not really compatible and has the same issues... now we look forward to do linear with redspace/redspace.... but here the contrasts are really hard as well...

Second big issue is the modulation in the blacks. The histograms are falling down in an angle in the shadows, that we have sometimes trouble to get a good contrast to the print..... highlights have a big headroom, but the blacks nearly nothing.... the shots are not underexposed, but looking like underexposed negative....

any thoughts ?

thx

rainer

 

[Sanjin Jukic]

Rainer,

forget use of RedSpace and Camera RGB/RedLog or CameraRGB/PDLog685 for output to "Cineon/DPX to print on film" is a way to go.

RedSpace is already "applied" look (LUT) to your R3D so you have to have a "clean/RAW look" and that is CameraRGB/REDLog or PDLog... to start your grade in whatever CC app.

Color Space: Camera RGB you'll find it at RedAlert or REDCINE apps.

Could be that you should start your R3D to Cineon/DPX conversion from the beginning!?

Also Chroma Denoise/De-Moiré turned OFF and OPLF Compensation turned OFF.

Then you can "apply" your CC look by using the LUTs provided from your post facility that can maintain more control of the final print image.

 

[Hans von Sonntag]

now we look forward to do linear with redspace/redspace.... but here the contrasts are really hard as well...

Hi Rainer,

I'm not a colourist but I have done a few successful filmouts which matched my expectations pretty well.

First of all I would leave the lin/log film LUT madness to the lab. You can screw up more than gain.

Secondly I would set the RA to 5000 Kelvin, 250 ASA and CameraRGB/RedLog. This wasy you can be sure that no information ist lost in the whites and the 12 Bit linear RAW data is lossless trasferred to 10 bit RGB. RedSpace applies a S-curve already you wat to create yourself to your custom needs.

Yes, the blacks in RedLog are "hovering" pretty high but it is easy for a colourist to bring them down as desired.

Hans

 

[Sanjin Jukic]

Secondly I would set the RA to 5000 Kelvin, 250 ASA and CameraRGB/RedLog.

Hans

Good points Hans and also to set

"Resample Filter" on "Lanczos" for getting a sharper picture as possible during the conversion process.

 

[Rainer Fritz]

Sanjin, first we differ between the color space and LUT. When you take camRGB as color space and look to this histogram, you will see that the chroma level is too low. Specially when you look to the shadows, there is not enough color information. When you then push the chroma on CC the colorspace is too flat and looks really bad. from my understanding the redspace LUT applies the S curve, not the redspace setting at the color space.
We tried camRGB/redlog and the LUTs at the labs dont match, because the files are not really log. I spoke with the colorist of slumdog millionaire and he graded three features shot on red, on different systems and labs, and he says the same, none of the log files which are coming out of redconverters are really log, its a gamma screwed linear RGB file.
Hans, im not applying any LUTs here, I'm speaking about the LUTs at the LABs. You need for filmout a real log file. When we apply the log LUT for the ARRI Laser and give it then on a red file, we see that this way is not possible.

 

[M Most]

We tried camRGB/redlog and the LUTs at the labs dont match, because the files are not really log. I spoke with the colorist of slumdog millionaire and he graded three features shot on red, on different systems and labs, and he says the same, none of the log files which are coming out of redconverters are really log, its a gamma screwed linear RGB file.
Hans, im not applying any LUTs here, I'm speaking about the LUTs at the LABs. You need for filmout a real log file. When we apply the log LUT for the ARRI Laser and give it then on a red file, we see that this way is not possible.

That's not quite accurate. Redlog most definitely is based on a log curve, it just doesn't happen to be Cineon Log, which is the specific curve you're talking about. The best and simplest emulation of a Cineon Log curve that I've found is obtained by using the PDLog985 gamma LUT with a 2 stop exposure reduction, as described quite extensively in Autodesk's white paper that has been referred to here in prior threads. This will give you a very good starting point for film work using conventionally derived film print preview LUTs, with essentially no information loss. Now, having said that, you can also use Redlog directly with film print preview LUT's, but you need to do some basic overall corrections to get the material to a sensible starting point. Another option is to convert everything to Linear Light and apply your own linear to Cineon Log conversion. If you really want to be clever, you can cascade a Cineon Log conversion LUT (or a curve correction to emulate that) and a film print preview LUT, and color correct below both of them, using the Linear Light data directly. This has some advantages in terms of mathematical accuracy and color grading "feel," but it requires a color corrector that will allow multiple cascaded LUTs in the pipeline, and the ability to work with 12 to 16 bit linear images to be truly effective.

 

[Rainer Fritz]

thx for the advise, i will try it with the pdlog985 -2 stops

 

[Sanjin Jukic]

Sanjin, first we differ between the color space and LUT. When you take camRGB as color space and look to this histogram, you will see that the chroma level is too low. Specially when you look to the shadows, there is not enough color information. When you then push the chroma on CC the colorspace is too flat and looks really bad. from my understanding the redspace LUT applies the S curve, not the redspace setting at the color space.
We tried camRGB/redlog and the LUTs at the labs dont match, because the files are not really log. I spoke with the colorist of slumdog millionaire and he graded three features shot on red, on different systems and labs, and he says the same, none of the log files which are coming out of redconverters are really log, its a gamma screwed linear RGB file.
Hans, im not applying any LUTs here, I'm speaking about the LUTs at the LABs. You need for filmout a real log file. When we apply the log LUT for the ARRI Laser and give it then on a red file, we see that this way is not possible.

Rainer,

as I already wrote above you can use "the combination" Camera RGB/PDLog... and then try out

to get in CC with your already "chosen" LUT that depends on the CC application you are using.

What you think that is "wrong" about R3D to Cineon/DPX file conversion you could ask and refer to Graeme Nattress!?

Also I'm a bit surprised about your CC questions here because I thought that you already know a lot of basics about R3D file "CC baking" :detective2: .

But now I'll give you a couple of tips:

1. Output each example frame from your project in DPX/Cineon (2K or 4K).


2. Drag and drop in Shake 4.1.


3. In Shake you have already TrueLight FilmLight plug in build in, add some others like Panalog from Orphanage

or use the KODAK 3D LUT Plug-in to apply a 3D LUT to an image in Shake.


4. Try out with TrueLight (set your parameters) and Panalog (Panalog2lin) as a Shake ready LUTs or import the LUTs for your specific CC app film out within Shake's KODAK plug in.

Some free LUT samples you could download from here.


5. You could "check" with your specific film out LUTs on R3D already on set/location using apps like

SpeedGradeOnSet or Gamma & Density 3cP (that I have) and output "look parameters" file ready for your specific CC app and Lab.


6. Also you can build your own LUTs (later your LAB ready) even on set/location with

Gamma & Density 3cP and The Light Illusion 3D Cube LUT Builder.


One more thing:

To get an idea about evaluation of CC file "baking" for the film out>>>

 

[Evangelos Achillopoulos]

We have successfully Film out two features with Log work flow the last three months.. the DPX's come from our partners pregraded, we did a straight printout no LUT's applied in our side... our partners where following our color management accreditation program and they used our custom LUT's (specific for their display devices not generic like shake truelight LUT's)... our lab is in Athens/Greece and they ware 5000 miles away (at least)...

In RED Cine we used CameraRGB/RedLog -1 in exposure.

For film out there is no magic LUT... that you drop a file and boom you have it...

It needs calibrated display environment up to date with custom LUT for the specific display device/filmrecorder/filmtype/emulsion combination.

And then the film recorder to be totally balanced and the developing lab to be extremely stable.

Only then you can have WYSIWYG in your film out...

Another solution is to grade in REC709 (linear) and expect something 75% close to what you see...

The above two films, that we film out, was graded in Apple Color... its not the grading software, its the LUT that make the difference...

 

[Sanjin Jukic]

We have successfully Film out two features with Log work flow the last three months.. the DPX's come from our partners pregraded, we did a straight printout no LUT's applied in our side... our partners where following our color management accreditation program and they used our custom LUT's (specific for their display devices not generic like shake truelight LUT's)... our lab is in Athens/Greece and they ware 5000 miles away (at least)...

In RED Cine we used CameraRGB/RedLog -1 in exposure.

For film out there is no magic LUT... that you drop a file and boom you have it...

It needs calibrated display environment up to date with custom LUT for the specific display device/filmrecorder/filmtype/emulsion combination.

And then the film recorder to be totally balanced and the developing lab to be extremely stable.

Only then you can have WYSIWYG in your film out...

Another solution is to grade in REC709 (linear) and expect something 75% close to what you see...

The above two films, that we film out, was graded in Apple Color... its not the grading software, its the LUT that make the difference...

Evangelos,

totally agree with you.

 

[Igor Ridanovic]

Rainer,

Autodesk published some measurements in an effort to understand what RED log options really do. The white paper they put out is inconclusive but it may shine some light on what RED curves are doing. The white paper came out late last year.

 

[Steve Shaw]

I've said this else where - many times.

Calibration 'print' LUTs care nothing about the source file, only the deliverable requirements.

They assume Cineon/dpx log as the working 'image data format'.

So, the best workflow is to use RedLog with all other metadata turned off (where possible), with ISO and Colour Temp set to optimum for the image, and same with noise reduction...

Then, with an accurate calibration LUT, grade through the LUT to give the desired result. This will 'force' the image into Cineon space.

Easy.

For some free LUTs, and more info see the website link in my sig.

Steve