Recent CML Tests
Canon C300 initial latitude tests
Epic exposed +/- 7 stops
With & Without HDR
Alexa +/- 7 stops
Infra Red exposure & Filter tests
Skintone tests
Alexa-Canon 7D-Panasonic 3700
Red-MX-SI2K
Digital Dailies and DI
Published : 11th August 2005
I've been hearing a lot recently about the issues facing DoPs and Studios
when using Digital Grading techniques for dailies review and Digital
Intermediate grading.
I well understand the problems that are being experienced, but am struggling to understand why they are happening in the first place. It is eminently possible to provide digital (HD or otherwise) dailies that are as accurate as film prints to the underlying captured image present on the negative, with the following DI process being just as accurate through it's use of the basic image data.
The problem is that not enough people yet understand this very simple workflow, and instead dailies are being creatively graded by telecine colourists, when that is in actuality the last thing that should be being done. Even telling a telecine colourist to 'grade in the middle' offers no guarantee that the image will be anything like what the DoP actually shot,
as I'm sure most DoPs can attest too.
This is because the creative process of telecine relies on the colourist to
control a system that has no datum point referencing the image actually contained on the original film. Telecines are not 'calibrated'. It's what a telecine suite exists for after all. Producing an image that is pleasing regardless of the underlying image information on the film being transferred.
However, there is a very simple process that can be applied to digital dailies, and the ensuing DI process, that ensures the digital image seen is always faithful to that captured by the DoP on-set.
Rather than using a telecine as a creative tool, it should be used to transfer an image via a fixed relationship between the film density and the digital data. This we have all been doing for years with film scanners,
which transfer images based on Kodak's Cineon film transfer characteristics, generating a known digital image file that represents the underlying film image very accurately, effectively creating a clone of the OCN. This image, usually a 10bit LOG DPX image, can then be shown via a calibrated LUT to present a very accurate 'preview' of the film image. In fact so accurate that the problem then becomes ensuring the film processing lab can print and process as accurately as the calibrated film transfer and LUT presented image.
The one caveat here is that the viewing medium, monitor, digital projector, etc., needs to be correctly calibrated as part of the LUT building process. Happily, companies such as Kodak provide systems to ensure that this is an easy process to complete, leaving the viewing conditions the only real variable. Dailies should always be reviewed in identical viewing conditions to those the final film will be viewed in, not someone's office with dodgy window blinds leaking light.
Therefore, the workflow that should be adopted for dailies is to transfer the film via a non-creative transfer engine (a telecine/real-time film scanner calibrated to Cineon/DPX density transfer characteristics) to generate an accurate LOG image which is a clone of the OCN. This can be in any 'video' or digital format you chose, from SD to HD, or greater, depending on the dailies viewing requirement and the 'telecine' used.
This 'cloned' LOG image can then be 'timed' via calibrated LUTs to present a very accurate viewing image, true to the DoPs work. It really is that simple, and I can't for the life of me understand why such a process isn't being demanded by DoPs and studios as the savings offered in understanding exactly what the captured image looks like are too great to be ignored. It's also a far simpler process than that presently used - relying on the un-calibrated creative process of a telecine and colourist.
Cintel have recently launched a 'telecine' (in reality a calibrated film transfer engine) called The dataMill which is aimed at this application. It's based on their Millennium telecine, but is a transfer engine that self-calibrates to the Cineon/DPX 10bit LOG image specification, locking to the film's D-Min and outputting an accurate 'clone' digital image in SD, HD 2K or 4K without any colourist involvement. I have also calibrated The
Spirit to output 10bit DPX LOG specification video, which is a more complex process but still not difficult to do.
This image data can be passed in real-time through calibrated LUTs to generate viewing dailies totally accurate to the OCN (or other film stocks)
being transferred, ideally as 25 across print images, or similar.
Simple control of the LUTs to mimic a timer's printer lights (+/- RGB,CMY)
can provide basic timing, without distorting the original captured image contrast, as is all too easy with normal telecine/colourist operation, providing total guarantee that the images being viewed as a digital dailies are totally truthful to the DoPs work.
The original Cineon/DPX LOG data can also be saved without LUT application for later DI grading, via systems such as Quantel's iQ, using the calibrated 'dailies' LUTs as an accurate guide to the look of the image signed off during dailies screening, ensuring a fully controlled and calibrated digital film workflow. Even if the OCN has to be re-scanned at higher resolution for the DI process, due to the dailies process being performed at too low resolution, the use of the same Cineon/DPX 10bit LOG transfer characteristics ensures the same LUTs will generate exactly the same look with the new high-resolution digital image data as seen during the dailies review as everything is calibrated to the same known datum.
Why would anyone not work this way, especially as the methodology is totally system agnostic and provides the image control and guarantees both DoP and Studio are looking for, at a cost point that is always going to be cheaper than via the traditional telecine/colourist route?
More information can be found at www.digitalpraxis,net
Best regards,
Steve Shaw
Digital Praxis Ltd
+44 (0)7765 400 908
www.digitalpraxis.net
Steve Shaw wrote:
> the use of the same Cineon/DPX 10bit LOG transfer characteristics >ensures the same LUTs will generate exactly the same look with the >new high-resolution digital image data as seen during the dailies >review as everything is calibrated to the same known datum.
This all sounds fine and dandy, but the truth is that different scans, even when done on the same equipment by the same facility, often differ, sometimes by small degrees, and sometimes by larger degrees. I have experienced this time and time again. Face it, film imaging is not an exact, repeatable science. Even when the same lab prints the same negative to the same print stock using the same lights, the prints can and will differ due to soup variations, as well as other possible variations (the lights themselves, the print stock, etc, etc.). Keeping these differences under control is the key. Quite frankly, accuracy and consistency through the post process is, to my mind, far more achievable
in the digital world than in the film world. The only area of the post process that has inaccuracies in digital imaging is the monitoring devices that are used.
Mike Most
VFX Supervisor
IATSE Local 600
Los Angeles
Steve Shaw wrote :
> I have also calibrated the Spirit to output 10bit DPX LOG specification
>video, which is a more complex process but still not difficult to do.
Steve, Is this directly out of the Spirit or using the 2K Davinci to modify
the HD output to log?
>Rather than using a telecine as a creative tool, it should be used to >transfer an image via a fixed relationship between the film density and >the digital data.
In theory, a "reference" transfer sounds like the way to go, but in practice
there are a lot of other factors involved. Something happens between wanting to provide a means to see what's really on the film and DPs egos. The jokegoes, "we color corrected the gray card and then color corrected the rest ofthe footage" just as we have always done, (dailies) and the DP who went tothe seminar and got hyped up on this technique smiles and says, "See, it works"
I'm sure you are familiar with various attempts over the years to provide RGB readouts for gray cards and such.
The use of D min, (the black unexposed frameline) has worked for scanners. It's a Spirit telecine dominated world right now. Is there a way to calibrate for D min on a Spirit? Has anyone used Firstlight from Davinci?
Dennis Politakis
Engineer
Matchframe Video
Burbank Ca.
Calibrating the Spirit to 10bit LOG can be done with the local control panel only - no da Vinci or Pandora. This is also true of balancing to D-Min. Its obviously how the Spirit works when in data mode via the Phantom Transfer Engine, you just need to use the same theory (but different controls) for the video output.
The beauty of using a Cineon/DPX transfer characteristic is that it only references D-Min, and then uses Kodaks fixed density to data transfer characteristics. This is easy to keep calibrated within tight tolerances - better than a film lab in most cases, as suggested by Mike. Moving between scanners (film transfer engines) using this approach has far better tolerances than having a neg. printed by different labs!
The reason I posted the initial message was in response to DoPs and Studios saying digital dailies were not as accurate as film prints, often for the reason stated above by Dennis (DoPs attempting to grade their OCN via the colourist). I'm just pointing out that digital dailies can actually be more accurate than film print dailies, to the underlying OCN image, if the workflow is done correctly. However, I know human nature is always going to get in the way with some people.
Steve Shaw
Digital Praxis
Steve Shaw wrote:
> The reason I posted the initial message was in response to DoPs and >Studios saying digital dailies were not as accurate as film prints, often >for the reason stated above by Dennis (DoPs attempting to grade their >OCN via the colourist).
I don't recall the initial message, but to me the common complaint about digital dailies is that they're not as accurate as film prints simply because they're not film prints. If your final release is on film, and you're not looking at electronic dailies in a calibrated digital projection environment, what you're seeing simply cannot be as accurate as what you would see if you projected film. Since most digital dailies are screened either in an office, in someone's hotel room, in someone's trailer, or in the back of the camera truck, calibrated projection is usually not available. This is without even considering the simple fact that monitor adjustments are changed all the time. Not to mention the significant difference in colorimetry between a typical standard or high definition video transfer and a film print, regardless of the care or skill of the daily colorist. For a television project, none of this matters because the final product is also electronic. For a feature, it does matter.
Although I applaud any efforts to make things more consistent, I reiterate that in my opinion, film is inherently inconsistent due to the various processes it's subjected to along the way. If you don't believe this, take two prints of practically any picture that were made on different days (even one day apart will do). Now try to intercut them. I can almost guarantee you'll see a difference from one to the other, particularly if you try to splice them mid-scene. That's simply the nature of a photochemical based medium. While it has many, many other advantages, complete consistency is not one of them. Ironically, it does happen to be one of the biggest advantages of digital imaging.
Mike Most
VFX Supervisor
IATSE Local 600
Los Angeles
I am aware of several facilities that do calibrated transfers. Two methods
dominate.
1/. For video transfers, the Kodak TCS system requires that the telecine is calibrated to a 21 step test film, so that it matches a digital target generated by the TCS box (external hardware). The TCS then reads keycode and applies a look up table calculated with Kodak film science. The result is an image that emulates a "standard print" and can be color corrected in the same pass, or recorded to disk and corrected later.
The system is attractive because the calibration ensures that even dailies from different machines are matched, because it needs no colorist intervention based on image content, exposure etc and because it captures more of the dynamic range than traditional 10 bit lin. Calibration can be done locally or with a da Vinci or Pandora system. Da Vinci first light automates the calibration process, but it has to be said that after the initial setup, the calibration does not vary much from week to week on modern telecines.
>Calibrating the Spirit to 10bit LOG can be done with the local control >panel only - no da Vinci or Pandora.
Steve could you expand on this - firstly because my understanding is that 10 bit log video direct from a Spirit is only derived from a 10 bit lin source
and does not have exactly the same benefits as a data transfer, and secondly because all telecine controls are remoted from da Vinci and I believe Pandora?
2/. The better way is to transfer 10 bit log data again using a calibrated telecine setting. 10 bit log data captures the entire neg dynamic range, and the calibration makes the transfer repeatable. In my experience, after
empirical testing, video from 10 bit log data is superior to a straight video transfer of any sort, ( the process above is the next best thing) The TCS is not yet available for data, but the same test film is useful. (Video settings are not compatible though) The concept is to map D Min to a digital equivalent and D max to its equivalent, but the use of Taf or TCS film to
match mid tones (for gamma tracking) is critical. The actual numbers can vary from facility to facility.
This process is ideal for DI (disk based) grading as Steve says. And I also agree that it must surely be the way of the future. Da Vinci, Pandora,
Quantel, Nucoda, Baselight, Discreet etc all have systems for this workflow
Reasons that calibrated transfers are not so dominant include :
- DoPs often have their input at the dailies transfer, not the final transfer. Clearly it makes sense to grade with the benefit of DoP input and record the result for reference later.
- Editors need to present a "good looking" offline for client approval. Calibrated transfers are not always good enough, especially if a "look" is
an integral part of the project
- The different workflow requires some investment in time and equipment, an investment that does not directly generate new revenue
These reasons do not prevent calibrated transfers, there are solutions, but I believe they slow down the acceptance rate.
Kevin Shaw consultant colorist
www.finalcolor.com
Laser Pacific's Leon Silverman has a "white paper" on the subject of "Digital Preview":
http://www.laserpacific.com/high_definition.html
Likewise, Laser Pacific's Emory Cohen and Terry Brown wrote some guidelines for digital post production that are on the same website.
Here is the Kodak "Telecine Tool Kit" :
http://www.kodak.com/country/US/en/motion/products/tools/ttk.shtml
And the Kodak Telecine Calibration System :
http://www.kodak.com/country/US/en/motion/support/tcs/index2.shtml
John Pytlak
EI Customer Technical Services
Eastman Kodak Company
http://www.kodak.com/go/motion
Recent Discussions
Advances in Volumetric Imaging
Rec 709 for theatrical release
Grading monitor calibration tool?