Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Best way to color correct RED footage

Ideal Color Correction

Ideal Color Correction

This is an interesting thread and full of good information. In my opinion, there are 2 ways to break this discussion down. Because this issue is so vast, I'm purposefully targeting bullet points from a more macro perspective. Hopefully some of this information will be helpful.

The first issue is "What is the ideal workflow for color correction?"
The second issue is "What is the ideal way in which to color correct?"

IDEAL WORKFLOW
Speaking technically about workflow might be a good start, because it's not as subjective.
Whichever path you explore, the ideal way to prepare Red footage for grading is commiting to a workflow that
A) requires the least amount of processing steps (ie: avoiding multiple passes of transcoding, even when uncompressed)
B) can perform the highest debayer quality
C) maintain the highest bit rate possible (R3Ds @ 12bbc native)

Each of the systems mentioned on this thread are affected in some way by these three basic criteria. Some systems support R3D natively whereas other require transcoding or base-band transferring. Some systems debayer in 2K full whereas others can do 4K full. And some systems can work in 10bbc whereas others float at 32.

The motive behind aligning your workflow to the highest common denominator is to maximize image potential in the actual grading process. The process of color correction is really a process of color separation. Precise qualifications can only be achieved with clean source material and the result of one's grade is partially dependent on the quality of one's source. With R3Ds, there is essentially a hierarchy of debayering potential which contributes to varying results when taking the same source media through different pipelines. For example, it is common to compare the differences between 444 and 422 processed footage. Though the naked eye cannot observe a difference, a computer surely can. Likewise, RGB footage characteristics will certainly behave better for a colorist, enabling the colorist to perform his/her job with less resistance from the media.

IDEAL GRADING
With respect to the opinions of filmmakers, enthusiasts and colorists alike, I estimate the colorist's talent makes up 2/3rds of an "ideal" grading situation. Regardless whether or not images are processed perfectly, the skillful eyes of a colorist will make all the difference in skin tones, balancing, mood and style. The balance hangs in the workflow leading up to the grade and the quality of the color correction equipment, monitor calibration and debayer quality.

COMPARISON
Because the world of R3Ds continues to evolve, so must the workflow in getting the images to perform better for the colorist. At PlasterCITY, we constantly test new workflows and compare the results in order to understand what each potential workflow offers.
Because many of the films we work on are mastered on 35mm, resolution, sharpness, color accuracy along with speed and list management play a key role in an efficient DI process. Below are two charts that assess debayering pattern results between different R3D supported tools. We always encourage people to do their own tests, but this is what we have found with the tools that we use in house.

This test was done with the best settings possible with each toolset. Since "best" settings obviously vary from system to system or plugin to plugin, it's not a true apples-to-apples test, but it does demonstrate results of unfiltered sharpening characteristics from a 4K source with familiar tools.

Below is a reference still from the film S. Darko (Donnie Darko 2).

My advice to everyone working towards improving workflow and color grading results is to test your available tools from start to finish against each other and time the results. There will inevitably be a cost vs. time trade off as well as a quality vs. investment trade off. Figure out where a project fits into this equation and you will develop finely tuned workflows that can aid in calculating your margin on investment.
 
Debayer

Debayer

Michael,

As always you take the time to post a comprehensive and informative response.

I sometimes forget how important the debayer process is and how different the results are from different engines and systems.

Thank you
 
Michael,

Wow! Thanks for the reference stills... what amazing differences. The Pablo stands out more to me (around the eyebrows)... is this because of the 4k full debayer?
 
Peter-
You are very welcome. It's my job to figure this stuff out. Sharing the information with others will hopefully improve this process for everyone. When it comes to post, we love a challenge.

Andrae-
In this comparison, 4K debayers were possible on the following tools:
Pablo Scratch RedCine Adobe RedAlert & RedRushes
2K debayers are the limit for
Color Log&Transfer Proxy & Native

So the vast majority support full 4K debayers but each manufacture does it slightly differently, thus the varied results.
One of the most interesting things we found in this test is that NATIVE looks very clean and sharp initially, but there is actually a significant amount of sharpening being performed and if you look closely (and on a scope) Native is actually the worst HD form of a full R3D transcode we could come up with.

One way this test may be very helpful is that we come across a lot of clients who do not need a full debayer in the mastering process. They want to transcode and just color correct the offline files to save time and money and the complexity of a reconform (which I think many people oversimplify). What this test shows you is what and how you should transcode for optimum results in mastering from a dailies source.

The biggest example of this is the output from Color. I know a lot of people are excited by this, including myself, but our results have shown that going back to R3Ds with Color and rendering 2K DPX frames yields poorer results than transcoding in RedRushes at half/high and grading the ProRes. You lose the Color room, but if you transcode dailies at half/high using RedLog, you will have a significant amount of dynamic range even in the ProRes to grade with. You give up the meta data and drop to 10bits, MPEG4.
 
Michael_PlasterCITY said:
One way this test may be very helpful is that we come across a lot of clients who do not need a full debayer in the mastering process. They want to transcode and just color correct the offline files to save time and money and the complexity of a reconform (which I think many people oversimplify). What this test shows you is what and how you should transcode for optimum results in mastering from a dailies source.

The biggest example of this is the output from Color. I know a lot of people are excited by this, including myself, but our results have shown that going back to R3Ds with Color and rendering 2K DPX frames yields poorer results than transcoding in RedRushes at half/high and grading the ProRes. You lose the Color room, but if you transcode dailies at half/high using RedLog, you will have a significant amount of dynamic range even in the ProRes to grade with. You give up the meta data and drop to 10bits, MPEG4.
Click to expand...

While I don't seriously disagree with this, there are a number of caveats that should be included when making the statement. First, you are talking only about debayering here, and while it does contribute a great deal to the overall quality of the final image, the fact remains that for color correction to be as flexible as it can be, and for the colorist to have the best opportunity to manipulate the image for a creative but clean result, it is very, very helpful to have all of the information from the sensor available without concatenated compression schemes and without nearly a good deal of the available value levels of the image format remaining essentially insignificant. So while the debayer result from, say, Color may have issues, the fact is that if the material is overexposed, a colorist will have more opportunity to recover information using that method than he/she will have using a "transcode to Prores" approach. So all of these things are tradeoffs, something that is understood by those with experience, but overlooked by those who don't have a lot of post experience. For those who will look at your statement as the "magic button" answer - the "just do this and everything is perfect" solution - there's more to it than that. Second, this is really only related to a do it yourself approach, because most facilities are already using full debayering (or, at the very least, Half/High) as part of their process, be it as part of the render path prior to color correction (as you have until this point for projects finished on Pablo) or as part of the finishing path (if one is using, say, Scratch as the color correction tool).

It should also be said that what you're advocating does not necessarily work directly in some of the situations you're describing, because most clients want dailies for editorial that at least to some degree represent the final visual intent. So most editorial dailies are transcoded using either Redspace or 709 gamma in order to retain that. Your comments regarding color correcting from Prores yielding superior results to the "R3d in Color" approach are dependent on transcoding using Redlog to retain as much of the original information as possible in a format that can be used for color correction. In most cases this would, at the very least, require a Crimson type of approach (i.e., a second transcode of included material), because most clients would not want to be cutting with Redlog coded images. Unless, of course, I'm misunderstanding what you're saying here....
 
Hey guys -I would like to chime in that technical challenges aside... I think the single most important point here is under Michael's IDEAL GRADING section.

I could not agree more, but think the percentage in reality, is more like 95% creative, 5% toolset as far as coloring goes.

Having had the luxury of working on a PABLO 4K, the difference between a Senior and Junior Colorist was FAR greater than the difference between the technology.

To me, the WORST thing about FCP Color, is that people off the street, suddenly assume that because they have a toolset, this makes them a colorist.

Even if it drastically improves what they shot. (Not to mention any real DP, is not a fan of a colorist, UNLESS, he is correcting an exposure mistake.) Most pro DP's I know, would rather capture the look IN CAMERA any day of the week. But I digress.

Historically, a COLORIST was a position that took YEARS AND more like DECADES to master in film. A lot of these dudes were in their 40's 50's and 60's before they perfected their craft with a few exceptions.

There is a lot more to it than correcting skin tones, or changing day for night. Secondary grading, vignetting, masking, etc.. techniques designed to punch a scene, or enhance a moment and on and on, is an ART FORM.

An art form that seldom few can "pick up" in a few sessions with Color or Scratch or even a powerful tool like Pablo.

I for one would be more worried about this aspect, than what the best tools and debayering techniques are... especially with how fast this stuff is changing.

I would LOVE to see senior colorists teach the ART side of it, more than the myriad of information about the technical challenges behind the art.

But then, they would most likely be putting themselves out of jobs :)
 
If its only 5% toolset... perhaps we should all rely on painters to color movies. Perhaps this ratio is best... 5% toolset... 45% mastery of that toolset... 50% aesthetics.
 
Hi Andrae...

45% mastering (i.e. acquired skill) and 50% Aesthetic (i.e. trained eye) kind of supports my argument of 95% art form ... don't you think?

id take a really good photoshop artist and give them a moving picture toolset
over a painter. less learning curve from medium to medium.

I am NOT a colorist... so my point is... id bet a seasoned colorist on FCP color, will kick my ass even if Im on a Pablo or Davinci. because in that context, my toolset and workflow advantage is irrelevant.
 
Michael, Great post, I did want to point out just in case the shot you posted wasn't a pickup, that we shot it with B15 4k 2:1, Not B16. Again, if the shot you posted is a pickup that was done later, then I apologize. I'm looking forward to working with you again on the winter project that starts shooting in a week and a half. Michael and everyone else at Plaster City was great to work with.
 
Tim Whitcomb said:
Hi Andrae...

45% mastering (i.e. acquired skill) and 50% Aesthetic (i.e. trained eye) kind of supports my argument of 95% art form ... don't you think?

id take a really good photoshop artist and give them a moving picture toolset
over a painter. less learning curve from medium to medium.

I am NOT a colorist... so my point is... id bet a seasoned colorist on FCP color, will kick my ass even if Im on a Pablo or Davinci. because in that context, my toolset and workflow advantage is irrelevant.
Click to expand...

45% mastering the toolset is way different than 45% mastering the aesthetic. The way I view it... a colorist is one part technician and one part artist. The ratio leans more heavily towards technician if the colorist does not have any creative freedom.
 
Steve-
You are right. The metadata header read the clip as B15. I had it mixed up with a different clip. Either way, the film is awesome and final DI looks better on the print than the DCP. -More on our findings about R3Ds prints to Vision later...
For now, this thread shows lots of good things to think about. The only thing I wish to reiterate is that these debayering results are one facility's findings. Everyone should do their own tests and evaluate the variables.
The subjectivity of talent is always going to be a point of great debate. Have at it. But this information can help people start their own workflows with their best foot forward. Admittedly, as Mike Most suggests, there are dozens of possibilities to explore with this information. Knowing where each path may take you will inevitably help narrow the path. And each narrowed path may turn out to be a different custom fit for every Red user.
 
You must log in or register to reply here.
Back
Top