Ask Mike Most Anything

[Marc Wielage]

The files that are usually needed are Avid MXF files (usually 1920x1080, DNxHD36), as the vast majority of TV shows are cut on Avid.

All the ones I've done lately have been either full-res DPX files or Avid DNxHD220x, which is great for 10-bit deliveries for TV. I've done some indie projects for DVD and Blu-ray release at the same res, and those have worked out fine.

Those who are interested in learning about the different Avid codecs can read this white paper:

http://www.avid.com/static/resources/US/documents/dnxhd.pdf

To me, DNxHD220x is essentially identical to Apple ProRes 444 HQ. All the major networks I know of have blessed these formats for post, but it's a good idea to check first to make sure the workflow agrees with what the network or channel wants.

If push came to shove, I would always prefer color-timing from a 2K 444 DPX file, and once the color-correction decisions are finalized, load in the 4K uncompressed files for the final render out. I don't see a need to monitor 4K 100% of the time, because of the huge performance hit you get on most color-correction systems. No doubt this will change over time, and eventually you'll be able to handle dual streams of 4K 3D with 20 layers without a hiccup. But not today.

Incidentally, this is much worse within the industry than it is in the consumer world. You've got colorists working on proper calibrated monitoring in proper environments, but you've got DP's and directors looking at dailies on everything from laptop computers in bright sunlight, to iPads, to video village monitors, to tiny CRT's in the camera truck, to home TV's that are set up to who-knows-what. And you've got studio execs looking at digital dailies that have been compressed beyond recognition and are playing in postage stamp sized windows on a computer screen and blown up to full screen size. The whole thing would be funny if it wasn't so ridiculous.

This is very sad, and very true. I continue to be flummoxed by directors and producers who are trying to make creative decisions about lighting and color based on what they see on iPhones and iPads, or even Apple computer displays. An editor/friend of mine (who should know better) was recently trying to tweak a picture in Final Cut Pro, and I asked him, "have you set the display gamma correctly?" He had that deer-in-the-headlights look. Not good.

 

[Marc Wielage]

After 6 years and about 40 full features and don't know how many shorts, using for chemical processes and projection Kodak Cinelabs Greece, we have achived a level of accuracy that is +/- 1 printer light... and that's a common ground for all my clients... If you allocate the 2 printer lights range in the density equivalent you can easily see from where the 3% number is coming...

Wow, that's remarkable. I mean no disrespect, but I've never seen a photochemical film lab that can guarantee prints at +/- 3 printer points, especially in a mass run. But that's here in America. Maybe things are better where you are.

Deluxe and Fotokem here in LA are very, very vague when you try to pin them down on how many points off a theatrical print can be before they consider it to be unacceptable. (Technicolor is no longer in the print business in North America.)

 

[David Mullen ASC]

I remember years ago finishing a movie photochemically at one of those labs, and the second answer print came back with an overall bias towards the pink and was also too bright -- the lab tried to say that this was within their normal tolerances! I tried to watch the print but after the first reel, I got tired of saying that every shot was too pink and too bright and I couldn't see the actual variations underneath anymore. They finally agreed to reprint it more accurately so I could give them accurate notes for the third answer print. But I was appalled that they considered this pink print to be acceptable.

 

[Evangelos Achillopoulos]

Wow, that's remarkable. I mean no disrespect, but I've never seen a photochemical film lab that can guarantee prints at +/- 3 printer points, especially in a mass run. But that's here in America. Maybe things are better where you are.

Deluxe and Fotokem here in LA are very, very vague when you try to pin them down on how many points off a theatrical print can be before they consider it to be unacceptable. (Technicolor is no longer in the print business in North America.)

Yehh +/- 3 is big time off... it makes a movie the "red movie" or the "green movie"... I explain it in a just before post, why in mass production they can't do it... apart from obvious production reasons...

The typical use of butterfly projections is to make that ONE or TWO copies super accurate no matter what is the mambo jumbo needed by the photochemical colorists... like PH tweak, temperatures and many many tricks, to make these two super copies and have a happy DoP and Director... the only diference is that magic is NOT on neg... so when the mass production starts all is like +6 printer lights miss match... different contrast etc...

To give an example... not even one ImageCare lab is on the list for US in Kodak web site... check your self...

http://motion.kodak.com/motion/Support/Kodak_Imagecare_Program/members.htm

Our lab in Greece is a Kodak Cinelab... and its the first that got the ImageCare in Europe... My film recorders are very well aligned with EVERY film emulsion EVERY time... and the color management I do is very strict...

All our corrections never going further than these 2 printer lights... and the result is accurate with NO extra tricks... some times even the first answer print goes to festivals... Like "Restoration" that went to Sundance 2011 and won the World cinema dramatic screenwriting award... Variety wrote "Marked by disquietingly beautiful imagery" it was neg development -> answer print -> festival...!!! http://www.motionfx.gr/client-work.html

How the hell you can live with 4 or 6 printer lights difference? is like night and day...

 

[M Most]

All the ones I've done lately have been either full-res DPX files or Avid DNxHD220x, which is great for 10-bit deliveries for TV. I've done some indie projects for DVD and Blu-ray release at the same res, and those have worked out fine.

My answer was based on transcoding for editorial, not finishing. I thought that's what David was asking, because I perhaps falsely assumed that for finishing, a conform would be done from the original media, which it often - but not always- is.

If push came to shove, I would always prefer color-timing from a 2K 444 DPX file, and once the color-correction decisions are finalized, load in the 4K uncompressed files for the final render out. I don't see a need to monitor 4K 100% of the time, because of the huge performance hit you get on most color-correction systems. No doubt this will change over time, and eventually you'll be able to handle dual streams of 4K 3D with 20 layers without a hiccup. But not today.

My answer was also based on television, not features, because that's what David was asking about. In television this season, the most common source format is ProRes 4444 because the Alexa is a rather dominant camera this year. I do realize this is RedUser, but even in the case of a Red originated show, you wouldn't be doing anything in 4K other than perhaps archival re-renders after the show is done and delivered - something we've done for a number of Warner Brothers shows.

 

[Florian Stadler]

Hi Mike,

I have a question regarding Log to Lin conversion in general/philosophically - RedlogFilm & AlexaLog:

I tend to grade straight from the log file and work mainly with Lift/Gain/Gamma controls in Davinci Resolve to first get the image in the ball park and then move on to a creative grade grade from there. I find that I get a good feel for the footage that way and am basically creating my own customized LUT that way compressing the highlights/shadows as dictated by the footage. Almost like in an old-school telecine session where you'd get back to looking at the "negative" when you wander too far away from the base grade and adjust the lights on the machine...

Am I missing something here? Am I missing something important in the Log to Lin conversion that would better be left to a LUT? If so what specifically would I be missing?

Thanks for this great thread...

 

[M Most]

Hi Mike,

I have a question regarding Log to Lin conversion in general/philosophically - RedlogFilm & AlexaLog:

I tend to grade straight from the log file and work mainly with Lift/Gain/Gamma controls in Davinci Resolve to first get the image in the ball park and then move on to a creative grade grade from there. I find that I get a good feel for the footage that way and am basically creating my own customized LUT that way compressing the highlights/shadows as dictated by the footage. Almost like in an old-school telecine session where you'd get back to looking at the "negative" when you wander too far away from the base grade and adjust the lights on the machine...

Am I missing something here? Am I missing something important in the Log to Lin conversion that would better be left to a LUT? If so what specifically would I be missing?

Thanks for this great thread...

Yes, you are missing a number of things.

First, understand that a log image is not an image, at least not in the sense of something that should be looked at and/or graded using traditional video style tools. It is, rather, a container, a way of fitting more information into a smaller space and using the available values in a more effective way for visual images. Think of it exactly the same way as you think of a film scan, because that's what these formats are. You would not likely look at a Cineon film scan and judge the image, and you would not likely try to grade it by using lift, gain, and gamma. In almost all cases, you would use a grading system that hopefully has log scaled controls (i.e., printer lights, or exposure/contrast controls) and observe the result through a LUT that is tuned to your delivery path and monitoring format. That is standard DI practice. The fact that your images are coming from a digital camera doesn't change the fact that they are essentially Cineon coded images. To get the most out of them - and by that, I mean to re-establish a proper greyscale and get the saturation correct - you should be using a grading path that consists of a base grade, followed by a Cineon Log (or LogC, basically the same thing) to Video LUT, followed by whatever color trims you want to do on the result. By going this route, you accomplish 3 things. First, you get to work with all of the information in the original image, because you're using that as your grading source. Second, you create a proper greyscale by using the LUT, and since you're doing a base grade prior to that, if the LUT is a bit "strong", you can change what's going into it to get back as much of the exposure extremes as you want, while still maintaining proper contrast. Third, you can do whatever final trims you want - usually some minor balance tweaks, and whatever secondary corrections you want to apply - in the video realm, after the basic image has been made correct. This gives you the richest and most detailed image possible.

As you noted, you can go without the LUT. Stefan Sonnenfeld often does. But - and this is not any kind of knock on whatever grading abilities you might have - you're not Stefan. Quite frankly, neither am I. And neither are most colorists. When you go without the LUT, you're usually forced to crush blacks in order to get proper overall contrast, losing detail and control in the process. You're essentially working the color corrector much harder than it should be worked. If you're crushing gammas as well, that sometimes leads to even worse results even if it feels right at the time. And even if you do it "right," you're not accounting for the fact that the log encoding puts the upper midrange values higher than they should be for a proper greyscale, so hot flesh tones tend to look very flat unless you crush those blacks even further. If you haven't tried using a processing path like the one I've described, you really should because it will get you much richer images, and much better results, much easier.

Now, although I've only mentioned a LUT, it should be noted that a curve grade can and does accomplish essentially the same thing, so if you want to design an S-shaped custom curve, that would do the same thing a proper Log to Video LUT does. There's also the issue of a color matrix, necessary to re-establish proper saturation and accurate color reproduction from any Bayer sensor camera. Red makes that easy by giving you Redcolor and Redcolor2. If you're coming off Alexa footage, you need to create a LUT that incorporates the matrix, for video, that would be their Rec709 matrix. I know that you're not the only one using the approach you're using, but I have to say it's not optimal, and until you try a more optimal path, you won't really know what you're missing. And although the direct approach can yield very dramatic results in the right hands, it's pretty difficult to get classicly beautiful images that way, especially if you value a proper, full greyscale and minimal black and white clipping. Unless you're Stefan, of course.

Try it, I think you'll like it.

 

[Marc Wielage]

I tend to grade straight from the log file and work mainly with Lift/Gain/Gamma controls in Davinci Resolve to first get the image in the ball park and then move on to a creative grade grade from there. I find that I get a good feel for the footage that way and am basically creating my own customized LUT that way compressing the highlights/shadows as dictated by the footage. Almost like in an old-school telecine session where you'd get back to looking at the "negative" when you wander too far away from the base grade and adjust the lights on the machine...

Our engineers at Technicolor (well, at least some of them) used to say: "the DaVinci (the old 2K) is essentially a variable LUT... if you can set up a basic look, and then do a final as another layer on top of that, it can work." I think this is where Stefan's philosophy comes in, and if you have his experience and reputation, you might be able to get that to work for you.

Today's more sophisticated LUTs will get you closer, and faster, especially if you're trying to finish in Rec709 color space. Me personally, I'd rather not waste a node or a layer just trying to simulate a LUT; I'd rather have a more accurate starting point, assuming the material is shot correctly.

Just starting off with the right Log -> Lin LUT can be very helpful. But so much depends on the nature of the original material, I'm not sure there's a one-size-fits-all solution. It's whatever works, based on the colorist's experience, how the footage was shot, and where the cinematographer wants to go with it. I've had LUTs so good, all we needed to do was minor tweaks here and there; I've had other cases where tweaking every other shot was an ordeal, like a self-root canal.

I've never had any luck with using custom curves to simulate a LUT... but then I'm not Mike Most or Steven Sonnenfeld.

 

[M Most]

Just starting off with the right Log -> Lin LUT can be very helpful. But so much depends on the nature of the original material, I'm not sure there's a one-size-fits-all solution. It's whatever works, based on the colorist's experience, how the footage was shot, and where the cinematographer wants to go with it. I've had LUTs so good, all we needed to do was minor tweaks here and there; I've had other cases where tweaking every other shot was an ordeal, like a self-root canal.

Root canal aside, I've sometimes experienced the same thing. But there was one very common thread to that experience - those LUTs almost invariably came from either a DIT or a dailies colorist. Those aren't the LUTs I'm talking about. In fact - and this is probably going to get me flamed by a lot of DITs here - I don't think those kind of LUTs should be used by colorists at all, other than purely for viewing purposes. A properly designed Cineon Log to Video LUT is not a variable proposition and it's not something that should be "tuned" to a specific shot. It's a mathematically derived conversion table that takes an image in a known Log format and converts it to properly display in Video space. That's it. It's not a color correction per se, and it's not a creative tool. It's no different than a video downconverter in that it takes a known source in, and spits a known result out. Now, the power of a properly designed LUT in a color pipeline is that the creative work is done by the colorist on the material as it was shot, not based on some unscientific result that was created without a proper pipeline in the first place. It gets you proper values based on what the camera captured, allowing you maximum flexibility but also giving you the most sensible starting point. There's no magic to a LUT, it's simply a mathematical conversion table designed to take one format and convert it to another. If you use a properly designed LUT on unencumbered camera original images, you'll get a good result. There is no ambiguity to this - it's based on the parameters of the color spaces and gamma curves that are used. Using DIT created LUTs invalidates that. If you're converting Red footage to RedLogFilm, and want to see a representative result of the photography in Video space, the easiest and perhaps best way to do that is to use a simple 1D LUT designed for the purpose (you can make one using, ironically, Arri's online LUT builder, by specifying LogC in, Video out, Extended range on both in and out, and no matrix). If you're shooting an Alexa, you can do the same thing, either with or without a matrix. In both cases you will get a result that's proper for the photography - no exceptions. If you substitute a LUT that incorporates some sort of creative intent - well, then you're on your own.

 

[Michael Olsen]

The following question comes from another thread, but I would love to hear the answer come from an imaging professional such as yourself, Mr. Most. It does, in some ways, pertain to recent posts regarding Log formats.

...I am wondering the advantages of [REDCODE] over Sony S-log. what people's opinions are over the matter, etc etc as I have shot Red before and I want to shoot Epic asap to see what I can learn about the RedCode capture codec vs s log.

As always, thank you and all the others for your time - this is a very enlightening thread.

 

[M Most]

The following question comes from another thread, but I would love to hear the answer come from an imaging professional such as yourself, Mr. Most. It does, in some ways, pertain to recent posts regarding Log formats.



As always, thank you and all the others for your time - this is a very enlightening thread.

They're really two different things. Redcode is a compression codec that is used to compress RAW data. It does not yield a viewable, correctable RGB image until you decode it, debayer the result, and push it through a color matrix and a gamma curve. S-log is a gamma curve and represents the end result of the 4 processes I just referred to in the Red work flow. So what you have to look at is S-log against the end result of the Red processes, not the beginning. In many cases today, for Red material you're often passing it through one of the Redcolor color matrices, and the RedlogFilm gamma curve. That gives you a color grading starting point that is, quite frankly, not that dissimilar from S-log (or Arri LogC, for that matter). The primary difference with Red is that you can control the metadata used for the debayer, thus better tailoring that RedlogFilm/Redcolor result as necessary given the original photography. That is what gives you the additional flexibility, not the result itself. So basically, the comparison would be S-log against the combination of RedlogFilm/Redcolor and the flexibility in creating that image. On that level, you have more control in Red's system than you do in Sony's, because you have no control over the debayer parameters in a Sony camera. And it's the same situation with Alexa unless you're recording Arri Raw.

 

[Michael Olsen]

:attention9ha:

Thanks for that, Mike. There's a lot of learning to be had here.

 

[paulherrin]

so that brings up the question, do you tend to mix different redcode color and gamma settings (redcolor1/2, redlogfilm, redgamma, etc) or pick one and roll with it? and how might that choice differ between commercial, television, and film. also, from both a practical and especially a subjective/creative viewpoint, any thoughts on their strengths/weaknesses?

does anyone know of any published information on these color/gamma spaces? official or otherwise. charts, tests, whitepapers etc, etc... maybe graeme can step in for us?

 

[Ian Vertovec]

I've had LUTs so good, all we needed to do was minor tweaks here and there; I've had other cases where tweaking every other shot was an ordeal, like a self-root canal.

My experience with Log -> Video LUTs tend toward the latter. For example, if I like the way the LUT rolls off the hilights but feel it crunches the blacks to early or too steeply, it can be very difficult to precisely pre-compensate the log encoded image. Instead I prefer to adjust the part of the custom S-curve within the grade that is creating the problem.

However, if you go with this approach Mike is correct in saying you need a really good curve grade that properly remaps your grayscale values. Otherwise your footage will never quite look natural.

 

[M Most]

My experience with Log -> Video LUTs tend toward the latter. For example, if I like the way the LUT rolls off the hilights but feel it crunches the blacks to early or too steeply, it can be very difficult to precisely pre-compensate the log encoded image. Instead I prefer to adjust the part of the custom S-curve within the grade that is creating the problem.

However, if you go with this approach Mike is correct in saying you need a really good curve grade that properly remaps your grayscale values. Otherwise your footage will never quite look natural.

I don't really disagree with that. I do feel, though, that for most people who don't do color grading every day for a living, and aren't really well schooled in exactly what the LUT or a curve grade is doing, using a LUT approach gets a very good result with very little pain. And, as you say, avoids the pitfalls of improperly designed curves. Having a LUT available is also useful in terms of being able to create a "proper" curve grade that then gives you the option of curve modification when necessary. Using a chart with known original and target values is one effective way of designing a proper curve grade, but many here probably don't either have such charts or know where to get them. For that group, a "mathematically correct" LUT is a good option. I know you use a curve grade for log conversion and, quite frankly, I do too much of the time. And it's nice to be able to tailor that conversion to the material, particularly in the lower end values. But for many if not most images, a LUT can work very well and presents a simpler solution for many.

 

[M Most]

so that brings up the question, do you tend to mix different redcode color and gamma settings (redcolor1/2, redlogfilm, redgamma, etc) or pick one and roll with it? and how might that choice differ between commercial, television, and film. also, from both a practical and especially a subjective/creative viewpoint, any thoughts on their strengths/weaknesses?

does anyone know of any published information on these color/gamma spaces? official or otherwise. charts, tests, whitepapers etc, etc... maybe graeme can step in for us?

What is true today is a bit different than it was a year ago. So my answer is based on what we have today.

For set viewing, a combination of either Redcolor or Redcolor2 and either Redgamma or Redgamma 2 is useful. Redcolor yields a bit more saturation than Redcolor2, so you can use either one depending on intent. Redgamma is a bit "crunchy," particularly in the highlights, but it yields an image for set monitoring that is probably closer to what is desired than Redgamma2, which is designed to hold more detail, particularly in the low end, at the expense of being a bit "flatter" in appearance. In other words, it's designed more for grading flexibility than it is final appearance. Now, in post, it's a different story. While Redgamma2 can work pretty well, you'll get a better greyscale interpretation and more grading latitude by using RedlogFilm. To me, RedlogFilm yields the best interpretation of Red's files of any of the alternatives, and the best greyscale interpretation since Red began. So I guess what I'm suggesting is to use RedlogFilm in post, and Redgamma for set monitoring.

Disclaimer: I don't shoot for a living. My concentration these days is in post production. And depending on creative intent, experience with log formats and log to video conversions, time available, and equipment available, your mileage may vary. Greatly.

 

[Marc Wielage]

For example, if I like the way the LUT rolls off the hilights but feel it crunches the blacks to early or too steeply, it can be very difficult to precisely pre-compensate the log encoded image. Instead I prefer to adjust the part of the custom S-curve within the grade that is creating the problem.

I think we're saying the same thing. I've used custom curves in Baselight to make minor tweaks to log scans I was dealing with, but they weren't doing the entire job of a LUT. I only used them to shift the gamma or black range to give me a little more wiggle room on the control and open up the whites a little bit -- merely a more "adaptable" base memory as a starting point. It's generally not something I would want to shift shot-to-shot, because there's just no time available for the session.

Haven't tried using custom curves yet in Resolve, beyond experimentation. I think the fact that everybody can work differently yet all wind up with decent pictures and happy clients shows you that there's a lot of ways to travel down that road.

BTW, I know a top colorist in town, A-list features, who works entirely in video mode -- doesn't color correct in Log at all. But they did a lot of workflow tests and tweak display and output LUTs to make this work. As he put it to me, "if what I see on the monitor gets me there, that's all that matters." The end justifies the means.

Me, I'll jump on whatever works. That was our approach on Shine a Light, and that was a completely custom LUT that was somewhere inbetween film colorspace and Rec709. Interesting adventure. Because this combined film scans and video, there was no other easy way to make it work. That was one of the easiest film -> Rec709 changeovers I've had, for the Blu-ray.

 

[Rob Ruffo]

Mike, I don't want to elaborate more, I'm measuring all the way around... moreover as we all know, there is nothing absolute in life all are relative and art is the most relative of all...

Art is also subjective... so to try to measure a relative and subjective value like art, its just waste of time... lets stop measuring tolerances and aim to have, happy artists... and that's what I do.-

i should not stir up a can of worms but I can't help it. I have a huge problem with people who make numeric claims (as you have) and then, when those numbers are questioned, say "forget numbers, it's all about art." There is problematic logical inconsistency there. Plus, reference projector calibration is first and foremost a technical science. It's not some kind of voodoo. There is a raw and hard technical side to what we all do. It's at the service of art, for sure, but art alone cannot make a good motion picture.

 

[Rob Ruffo]

On a more positive note, thank you so much for such an amazing thread! Wow!

And a question: Redcolor 2 seems to less often yield yellowish skin tones than Red Color 1 - is that my imagination?

Another question: Red Color 2 seems to be able to to eb pushed into very high saturation (Saturation parameter set very high) without channel clipping to a greater degree than RedColor 1, even though at 0 it starts as less saturated - so in the end, it seems, that RedColor 2 is better able to achieve "clean" high sat images than Redcolor 1, because it can be further cranked up without issue - is that true? AM I mistaken?

Thank you again.

 

[M Most]

And a question: Redcolor 2 seems to less often yield yellowish skin tones than Red Color 1 - is that my imagination?

Another question: Red Color 2 seems to be able to to eb pushed into very high saturation (Saturation parameter set very high) without channel clipping to a greater degree than RedColor 1, even though at 0 it starts as less saturated - so in the end, it seems, that RedColor 2 is better able to achieve "clean" high sat images than Redcolor 1, because it can be further cranked up without issue - is that true? AM I mistaken?

It sounds like you're referring very specifically to behavior in RedcineX. That represents only one set of circumstances. Many other programs can be and are used to directly access R3D files, and many do not behave in the way you've described. In my view, increasing saturation in a post grading environment is usually not a good idea as it increases noise, particularly in red tones. Many, if not most, colorists would much rather decrease saturation than increase it for that reason. There is some debate on this issue, but in either case, you cannot use the characteristics of one program - particularly one that is working on the debayer parameters rather than post processing - as a basis for general assumptions about image processing, regardless of the camera.