Mix and match Log3G10 LUTs in Resolve (Primers, philmColor, IPP2, etc)

[Michael Tiemann]

I am finding that my clients really love the Omeno Primers, which is wonderful! But color is very subjective, and strangely I prefer RED's IPP2 LUTs as a starting point. I also really like philmColor for getting looks going. All this to say: I've tried to make it as easy as possible to swap out one group of LUTs for another with the least amount of disruption to my overall workflow and projects. And I think I've discovered a few things that might be helpful to others along the way. These findings are not authoritative, just what I've learned thus far; I'm also always happy to receive correction and/or constructive feedback.

BACKGROUND

Davinci Resolve qualifiers are really made to work with REC709. Wider gamuts (especially RWG) and flatter profiles (especially Log3G10) cause the qualifier to wildly over-shoot intended selections. And it doesn't appear to be on the roadmap of Davinci Resolve 16 to change that fact. Which itself is not really a problem: Resolve's 32bit floating-point color is more than capable of mathematically inverting the transformation from RWG/Log3G10 to REC709 and back again with essentially no generational loss. For this reason, REC709 is not only a perfectly fine colorspace and gamma for timelines, but it may well be the best setting for general Resolve use.

MIXING and MATCHING PROBLEM

The LUTs I use all take RWG/Log3G10 as input. The Omeno Primers appear to produce P3-D65/Gamma 2.6 output. Resolve has a mathematical transform that can convert that to REC709 in a lossless fashion. The IPP2 v1.13 LUTs target a wide range of color spaces and gammas, including P3-D65/Gamma 2.6 (matching Omeno Primers) and REC709. But there's a twist.

The Base primer yields a significantly darker image than the IPP2 LUTs. I don't know why something middle-gray-ish to IPP2 gets mapped 25 IRE less when going through Base, but it does. Which makes the mix-and-match exercise much more than a simple A/B color choice. Grouping together two identical clips of a CDM color chart and Using Resolve's Multiviewer to view the two clips side-by-side, I used the waveform monitor to analyze the difference between the two.

EQUALIZING LUT GAIN

It may be obvious to many who have spent a lot of time color grading, but using Gain on a gamma-corrected source has the effect of increasing contrast (because higher IRE values are gained up proportionally more than lower IRE values). Applying gain to make the peak brightness match, then applying (negative) lift to make the black tones agree tends to increase contrast even more. Changing the ISO on a per-clip basis does correct the discrepancy of luminance response, but it's a total nightmare to have to manually change the ISO on all the clips when you want to swap between one LUT and another. The solution I have found to this problem is to create an ISO Gain node before the Primer LUT node and to set the ISO Gain's Gamma setting to Linear (not Use Timeline). When using Gain in Linear space, the effect back in Gamma space is precisely that of increasing/decreasing exposure. All contrast values remain in the same relationship to one another.

EQUALIZING LUT CONTRAST

The Omeno Primers come in four video responses (Base, NTR, Vivid, and Punch). The IPP2 LUTs used to come in four video responses with 4-5 highlight rolloffs, which is 20 choices per colorspace/gamma output. With some trial and error, and finding the right gain correction above, I found overall good matches as follows:

Base: ISO Gain 1.45 matches IPP2 LOW CONTRAST, SOFT HIGHLIGHTS
NTR: ISO Gain 1.36 matches IPP2 MED CONTRAST, MED HIGHLIGHTS
Vivid: ISO Gain 1.27 matches IPP2 HIGH CONTRAST, SOFT HIGHLIGHTS

Punch gives even more contrast than IPP2 HIGH CONTRAST

MIXING AND MATCHING MORE LUTS

Phil Holland makes a big deal that most of his LUTs are luminance-preserving. Omeno makes a big deal that Primers should be the first thing in the chain after ISO and White Balance. But philmColor GHOST stocks are designed to be very light treatments, more like filters over the lenses than treatments to be applied in the chemistry/timing. They can take RWG/Log3G10 RAW output and pass it along. The philmColor BASE stocks are probably better to use after Primers, which means that Primer outputs must be converted from P3-D65/Gamma 2.6 to RWG/Log3G10, and then back again after the BASE stock is applied. Again, Resolve's 32bit floating-point math in the OpenFX color conversion nodes makes this a snap.

I've bundled up a project that includes the test chart and Resolve node trees. You need to have your own LUT files. The CDM chart was shot with tungsten light balanced closer to daylight (plus some ambient daylight) on a DSMC2 Helium S35 camera. You can access the DRP file and media here:

https://www.dropbox.com/sh/4qsb9wvtjru8g65/AABx5KETlalvZYKtaa1wDU5Da?dl=0

Have fun, and please send me corrections to any errors I may have made. Thanks!

 

[Marc Wielage]

I casually know Hrvoje Simic over at Omeno Design, and I think the LUTs are OK (and usable), but as with all this stuff you can duplicate it all if you have a good colorist who knows what they're doing. You can get these same looks in Baselight and Lustre and Nucoda and Resolve and everything out there. It's not necessary to use a LUT to get reasonable images from well-exposed Red material. I do like using IPP2, RWG and 3G10, but the rest I just dial in. It's not rocket science, and once you have that look, you can keep it going throughout the whole project. I don't have a problem with anybody who wants to use them as temp looks, but trust me, a good colorist can do the exact same thing.

The advantage of doing it manually is you can customize it on a shot-by-shot basis and compensate if and when the exposure goes off the rails, or if the director & DP want a vastly different look. LUTs are not always the answer. I do agree that there are very tweaky, complex custom LUTs created for a specific project that can work in certain situations, but I generally avoid using them as "normalizing" LUTs early on. (The one exception for me is SLog-3, which is a total pain in the ass.)

 

[Hrvoje Simic]

Whoa. More workflow info was planned but now Michael made me do it sooner.

LUT is body which associates to known usages, logic and outcome.

Primers go beyond those.
They are based on image philosophy, defining the whole pipeline from camera analysis to signal shaping.



I'll add more info here when the time allows it.

 

[Marc Wielage]

The science behind what Hrvoje does is sound, and I appreciate the fact that he's using real charts and good real-world conditions. But I'm always skeptical of a LUT being any more than a reasonable temporary look on set. I do know DPs who make custom LUTs for different applications, but it's a case-by-case basis. And even then, one LUT is not always going to work for every shot in the show. A colorist can make vast adjustments when necessary to get all the material in the same visual place, simply because the knobs move.

Bear in mind that a big part of it often involves adjust level or color differently in different parts of the frame, and you'd be surprised how often that happens. A LUT can't do that, and a LUT also is going to have issues the moment you jump into ACES or a radically different color space. That's a natural limitation of LUTs for very real mathematical reasons. I do think LUTs (especially technical LUTs) have their uses, and I'd always rather the people on set and the editor look at the footage with some reasonable kind of look that compensates for Log space. And I still encounter newcomers who try to edit in Log and wind up getting used to that uncorrected look, which leads to lots of stress in post.

 

[Hrvoje Simic]

But I'm always skeptical of a LUT being any more than a reasonable temporary look on set.

Understandable, from your current perspective I'd be also.

I'll go through the logic which helps to understand this philosophy and reaching different conclusions. Based on years of interdisciplinary research from acquisition to post and repeatable results from the developments.
Formulating that into a meaningful form takes a significant chunk of time I don't momentarily have, but soon enough...


Also, Michael - I do not recommend mixing/stacking tools like these. They are far too specific and depend on input signal.

Phil likely designed his tools based on specific workflow logic and predictable signal input from base Red colour science as well, and although I understand and respect the exploratory nature behind the notion, this route leads to unpredictable results. They will not do the same thing if the input differs. In any case I don't recommend mixing LUTs, too much variables which can mess up the signal.
Whether it is specific stock/look design as with Phil's tools or neutral normalized image basis with Primers,
past that point it is best having specific manual control over the image.

 

[Marc Wielage]

Also, Michael - I do not recommend mixing/stacking tools like these. They are far too specific and depend on input signal.

Yes, very true. A couple of months ago, I got a panicked call from a client who had a DP who tried to color-correct his own feature, and it went right off a cliff. I was pulled in to try to salvage it, and discovered that his first two nodes were both different LUTs in a row (!!!), and of course one clipped images and the second tried to compensate for that. Eliminating both LUTs gave me the room to fix it quite a bit.

I've seen Phil's LUTs as well and I think they can be useful in terms of giving filmmakers and DPs inspiration for new and different looks, but I'm as suspicious as any of this stuff in that a great colorist on Baselight or Lustre or Mistika or Resolve is going to be able to match that look and not cause as much damage. And it's especially hard when you get a project that has to deliver in 1000 & 4000-nit Dolby Vision, P3 XYZ theatrical, and 100-nit Rec709 home video.

 

[Phil Holland]

I've seen Phil's LUTs as well and I think they can be useful in terms of giving filmmakers and DPs inspiration for new and different looks, but I'm as suspicious as any of this stuff in that a great colorist on Baselight or Lustre or Mistika or Resolve is going to be able to match that look and not cause as much damage. And it's especially hard when you get a project that has to deliver in 1000 & 4000-nit Dolby Vision, P3 XYZ theatrical, and 100-nit Rec709 home video.

It's not really about replacing a colorist or anything like that. I just ran into about two dozen folks here at NAB and it's good to listen how and where my Creative Cube LUTs come into their workflow. For the studios, commercials, and filmmakers they get used in a variety of ways and they certainly put things into a specific zone for those jobs that don't have the budget for a week or month of color work. A good number of people out there find a handful of useful ones to toss into their camera just for monitoring purposes even to bring a smile to the puckered lips of a grumpy director.

A good colorist can do anything, I'm mostly giving some options on where to start and consistently providing a look across the board if desired. As mentioned, the ghostStocks indeed operate more like Color Science and they feature subtle transforms that moderately operate like the evolution of various film stocks in creating a unique look. As these cameras are getting better and providing more color informatio from Dragon onwards to Monstro, Helium, Gemini, and Dragon-X it's about guiding a sensible pleasing look and directing a simplified and refined palette as well as tones.

Remember, philmColor adheres to the RED IPP2 workflow and goes into the Grading Zone of the pipeline. As always a little graphical help:
http://www.reduser.net/forum/showthread.php?161964-RED-IPP2-Introduction

philmColor is always designed to be used as such: REDCODE RAW (or ProRes) > Grading Zone (curves, CDL, etc, philmColor) > Output Transform (Output Tone Map/Highlight Roll-Off and your Output Color and Gamma Space (like Rec.709/Bt.1886).

I've done my best for a long time to keep the vast majority of the Creative Cubes as tonally neutral as possible to respect individual filmmakers ways of exposing and working. There's a few in there, like the Bleach Bypass related looks, that will obviously do things related to tonal pushing related to the core of what inspired the LUT in terms of the process. These have been tested with all of the different Output Tone Map and Highlight Roll-Off variations and literally with thousands of shots from many RED owners, productions, and even from RED.

I've received some interesting requests since the last update that I have been working on that will have their own little world, which will be clearer shortly. Though I don't totally recommend some people have discovered you can indeed "stack" philmColor Creative Cubes for even more insanity and I've now seen them used on ARRI LogC material before their desired output transforms on a few shows. Sneaky guys out there breaking the rules.

 

[Hrvoje Simic]

Yes, very true. A couple of months ago, I got a panicked call from a client who had a DP who tried to color-correct his own feature, and it went right off a cliff. I was pulled in to try to salvage it, and discovered that his first two nodes were both different LUTs in a row (!!!), and of course one clipped images and the second tried to compensate for that. Eliminating both LUTs gave me the room to fix it quite a bit.

Yup, that is a problem. Epecially when one doesn't know what the LUT does and how it can affect the signal values.
When one knows exactly what it does and uses it properly, and when it is properly designed , there is a huge potential.

LUT is just a lattice based container for the transformation. There is no "good" or "bad" in the container itself, it can only be in what it contains. Dismisssing a LUT is like dismissing "mp4".

That's why I'm avoiding talking about the LUTs in the context of Primers. Too much preconceptions baggage in a domain saturated with all kinds of stuff. Forget about the LUT, and associations from past experiences, "hollywood blockbuster insta-colorist" type of stuff. It's what's inside that matters.

I've seen Phil's LUTs as well and I think they can be useful in terms of giving filmmakers and DPs inspiration for new and different looks, but I'm as suspicious as any of this stuff in that a great colorist on Baselight or Lustre or Mistika or Resolve is going to be able to match that look and not cause as much damage.

That depends on what is acheived with this type of transformation.

Manual work can also cause damage. That part goes both ways. LUT can be more precise, stable and consistent than manual work.
Every operator affects the signal in multiple ways, depending not just on properties of the operator but also on properties of used colour space and gamma. There is no free lunch. You manipulate one thing, vectors for example, tonal transition smoothness will change, or microcontrast, or specific colour luminance, etc. Vectors will have one types of problems and limitations in log, another in linear, another in gamma corrected image. Or saturation, contrast, etc. It can go haywire easily and variables multiply. LUT is more specific, and can target specific tone on Wednesday, without messing up other things. It all depends.

Phil has likely done extensive work as well on analysis, design and tests and that is not something which can be compensated just like that with a few quick moves in post. You may imagine months and months of work on a single transformation which does not lose value just because it's finally applied with one click in post. Value is relative to perception and personal system of values. There is huge work involved to get to that transformation step. If it is achieved with extensive, systematic and tested workflow it is a legitimate tool, which expands creative options.

As far as matching the "look"...one can match the look per shot/scene, per properties which the tools can affect and per visibility of manifested results. No other reference whatsoever, no in-depth insight of the camera performance itself - the very basis of the incoming image, no in-depth insight on the full effect of the LUT in other circumstances and no guarantees whatsover the same look done manually will work on other scenes, in which case more work is required to get to the same (similar) result. There is this data-centric fantasy that matching ideal values (colour matrix for example) and rudimentary targets will do the same thing with all cameras. As if camera processing and sensor properties = their ideal mapped targets. Like it's that simple. That comes from people who have never actually used them or haven't used them long enough. Also - matching is the attempt of re-creating something already done by someone else. None of what Phil or me are doing is there to compensate or undervalue the work of a colourist, it is there to empower the workflow, with differing approach, purpose and logic. You may think of it as expansion, not replacement.

I started making these for our post initially, so the first light is a joy instead of PITA in many cases. First started with orange people with first SGamut in 2015. Now I'm not doing any post without Primers if the material is shot on cameras I've analysed and designed Primers for, and there is no frikking way I want to shape the image from scratch now, or let someone else decide about the initial transformation, knowing exactly what these do per each camera, what they achieve, and what it took at what levels of complexity to get to this.

Let me consider how much I care about IDTs and 45638464 steps of signal gymnastics and image variables after thousands of hours of camera analysis, signal shaping and tests for a single, scene-independent, stable and reliable transformation step normalizing the sensor-specific imagery, which I can carry with me and can apply in camera, recorder, monitor, NLE, grading app or stills app:
..
...
....considered.

 

[Hrvoje Simic]

Beauty and Colour


There is objective and subjective part of beauty and aesthetics.

Objective part is universal and depends on principles of harmony, proportions, correspondence and stability.
Subjective is particular and related to taste and conditioning.

Primers are focused on the objective part.




DIGITAL & CELLULOID


Digital cameras are imperfect. They are imperfect in a different way to film.
Digital quantization and signal mapping, especially into log gammas and wide gamuts introduces more specific imperfections. The deviations are neither pleasing as with film nor cannot be compensated the same way as with film. Modifying the digitally acquired image on top of digitally acquired image’s signal deviations introduces further unnatural offsets of the signal (compared to film) which is why:

a) Digital can often look as “washed out” compared to film or "souless" or “fake” or “electric” or any other association which comes from that “digital feel”.
Which entails complex work just to get closer to image basis which grades and finally looks closer to film.
b) As soon as scene light or post look goes off white balance new problems emerge, weird tones pop out, often difficult to remove and image aesthetic differs more to film graded off white balance

Great colorists can still make the digital image look fantastic but nevertheless they are forced to continuously start on a compromise as a basis.
Compromise which wasn’t there with film and its organic and more gradual offsets.

This is where Primers come in.



PRIMERS PHILOSOPHY


Primers are image shaping tools designed to optimize digitally acquired imagery.

Their design is based on thorough camera analysis, with custom developed tools and workflow and mutiple levels of testing.
They take the original camera input, mapped into camera maker's base colour science and transform it into an optimized image.
Primers EQX take the unmodified log input, Primers Standard series are designed for compact cameras with limited bit depth and require in-camera settings mod before the signal get compressed.

They allow visually referent monitoring and top quality first light. They remove chances for underexposure, reduce needs for NR and complex compromises in post.

1) They are designed for those DoPs who had “ I wish I could capture this the way I see it” moment in their lives.
The only main difference to that intuitive notion is the technical limitation of mapping bright high saturated values, which are designed more subtle to allow natural and filmic offset to signal extremes, and healthy signal which grades well. Camera sensor and processing limitations are minimal problem these days.

2) They are designed for DITs to make their lives easier and more fun, and to make DoP, director and clients happy, without wasting time on unneccessary signal gymnastics and fiddling while cards are coming, just to get an accurate quality image.

3) The are designed for colourists who want to focus on creative look with stable image basis and don’t want to waste time, focus and energy every single time to fix something already fixed.

The term “optimization” refers to:

1) Adaptation - achieving optimal signal mapping for sensor performance based on original camera maker’s colour science (affecting density, contrast, micro contrast, richness, separation, DR utilization, lattittude utilization, noise and texture properties etc)
2) Stabilization - achieving tonal and texture consistency across the exposure range and shooting scenarios
3) Harmonization - achieving optimal tonal relations
4) Normalization - achieving technically correct and reliable signal for post production
5) Naturalization - achieving visual reference and organicity and removing the “digital” feel from digitally acquired image

These five categories are formulated just for clarification, they are not consecutive nor separated in the process but inter-related, with enormous amount of variables to juggle and exceptionally complex to meet all goals in the same time, which is why this process takes months and r&d to get there took years with multiple revisions of the whole workflow and final outcomes.

No color bias. Colour neutral.
Since cameras vary in spectrum coverage and dynamics, judgement calls are made to keep the prioritized values consistent while assuring image’s overall visual neutrality and organicity.




PRIMERS SET


Primers EQX sets consists of 5 Primers

01 Base
02 NTR
03 Vivid
04 Punch
05 DMax


- Base and NTR are designed primarily for colourists and allow most creative shaping range. Base is most flat.
- NTR is the golden middle for usability in monitoring and post. Not as flat as Base but still having most texture visibility over the exposure range with more shaping room in shadows and highlights than the following 3. If one wants to see most of usable DR on set - Base and NTR are advised.
- Vivid and Punch are designed mostly for clients, directors and set due to most contrast. Also, make up, hair, art dept. and wardrobe can benefit with more popping out. Punch has highest advisable contrast for overall DR utilization, healthy exposure decisions and colourist having nice thoughts of you after reviewing the material.
- DMax is most dense, a bit darker as a result and has lowest noise floor per contrast. Highlights are close to NTR. Naming is hommage to film, not exact analogy. If you are shooting low light and unsure, use DMax for monitoring.

All are coloristically the same and designed in subtle yet practical increments with special contrast response design in each for optimal density, filmic midtone contrast, best noise floor and texture per DR utilization, and harmonious tonal distribution, while having graceful transition to signal extremes.

For the DoPs additional importance:

For the best image quality:
- Do NOT expect some miraculous detail popping out of highlights which you don't see with Primers
- Do NOT expect for the guy in post to dig more juice out of the shadows which you don't see with Primers on set
It is either not there or pushing it to get it will cost you.




PRIMERS WORFKLOW


Primers are sensor-referred signal shaping tools.

That approach makes them scene-independent.

They work in EVERY situation and optimize the image regardless of the light type, exposure, ND tint or scene colour bias.
In the context of post production they are designed to be first light tools, which make first light process a) practical, b) intuitive, c) consistent and d) high quality
For the best result one has to white balance the image pre-LUT, while Primer is applied, and Primers will optimize the result.


Signal chain

1) RAW settings - temp/tint (tonal signal positioning), ISO equivalent (luminance signal positioning)
2) Primer - in Resolve it goes on the first node
3) Creative look shaping goes after the signal is optimized


- NO other transforms prior or on top of Primers. They have “falloffs” and a whole bunch of other things all implemented inside.
- NO saturation, “exposure”, contrast of any type, “shadows/highlights” or qualifiers pre-LUT
- NO auto chart matching or "auto anything" for that matter - because
a) that is fundamentally flawed,
b) original colour relations have to be intact as coming from the sensor
c) you will get superior results with manual temp/tint teamed with vectorscope insight and Primers
d) Primers already know what the sensor and base colour science do EXACTLY and what to correct EXACTLY, MORE SPECIFICALLY and MORE PRECISELY than eyeballing, so anyone color correcting has to compensate the light variable, Primers do the their stuff on camera induced variables.

In most cases temp/tint sliders are sufficient to get a proper tonal balance. The need for trackballs for basic CC is more rare. Subtle WB adjustments make the world of difference in base image quality and further cumulative image manipulation.

Trackball creative toning and similar creative approach segmented over luminance range should be done after the image is optimized by the Primer, after which creative tonal offsets become more stable and organic looking. The same logic applies on other tonal particulars. First the technical normalization, then the creative work on stabilized material. That way the chromatic circus of digital is gone and the image starts to behave more naturally for look design.

 

[Martin Konrad]

I like the approach to start with a sensor specific Primer and adding all creative decisions in the grade to a perfectly repeatable, neutral starting point.
The examples on your website are great, I like the skin to be a bit less yellow than the standard Ipp2. But examples are all Dragon, can we see somewhere
your what primers do to Gemini?
I also have a Davinci question, are multiple nodes pulling the grade in opposite directions degrading the image (assuming everything is done in RAW)?
If I were to apply a Node adding contrast and a warmer whitebalence (via gamma) and at the end I add another node loosing a bit of contrast is the image
quality less than going back to the original node and adjust? Hope I make sense...

 

[Hrvoje Simic]

The examples on your website are great, I like the skin to be a bit less yellow than the standard Ipp2. But examples are all Dragon, can we see somewhere
your what primers do to Gemini?

Those are examples from the previous generation. It gets better. I'll post some new ones including Gemini. Main tones is the first thing you typically notice, there is more in specific tonal relations, separation and definition. Also..there is much more than that..but is better first seen than said in advance.

I also have a Davinci question, are multiple nodes pulling the grade in opposite directions degrading the image (assuming everything is done in RAW)?

If I were to apply a Node adding contrast and a warmer whitebalence (via gamma) and at the end I add another node loosing a bit of contrast is the image
quality less than going back to the original node and adjust? Hope I make sense...

Once the image is normalized creative manipulation yields more stable results.
Primers have contrast starting points defined, and on multiple image parameters. It is not just "any" contrast. After that it is up to you.

Generally - it depends on what exactly you are doing, image manipulation is a destructive process, just depends when things become perceivable. If it is exactly the same type and application of contrast it is better to reduce the cause then compensate it with another process.

 

[Marc Wielage]

I also have a Davinci question, are multiple nodes pulling the grade in opposite directions degrading the image (assuming everything is done in RAW)? If I were to apply a Node adding contrast and a warmer whitebalence (via gamma) and at the end I add another node losing a bit of contrast is the image quality less than going back to the original node and adjust? Hope I make sense...

I think for color, you have to develop what I call a "node philosophy" (or layers for those systems that use layers), where you start with a basic image, balance it to something approaching "normal" in your specific color space, and then start adjusting it from there. I prefer to do all this manually for the simple reason that I can tailor that look for the specific way the image was shot: underexposed, mixed lighting, a specific color temperature, overexposed, or even in a way that relies on me to relight the scene if necessary. A LUT alone can't do that.

It is true that if you slam the picture one way early in the chain -- say, crushing or clipping it to the point of distortion -- even in a 32-bit float system like Resolve, you will wind up with a distorted image that you can't completely recover. I see neophyte colorists who have fallen on bad habits over time, and they don't understand why they have to work so hard to get good images... because they're screwing up the image early on. Destructive LUTs can do this. Bad Raw settings can also do it.

There are several colorists in LA that have an interesting philosophy that they do 90% of their work entirely in Lift (blacks) and Gain (whites), and it's true that you can get in trouble with midtone (gamma) adjustment in certain cases, particularly if you're trying to increase it. That's a sure way to introduce noise, particularly in Resolve. Understanding custom curves and getting the right balance is critical.

As to LUTs, I'd point to Juan Melara's excellent tutorial where he reveals how he duplicated a $600 LUT just with the controls in Resolve:

Granted, a lot of work was involved with this, but it's fair to say there are several different approaches you can use, and they work on pretty much any LUT in the world (with rare exceptions). I know of many colorists who only use the initial temp grade (done with a LUT) as a general guide, and then they just match it manually for the final. Sometimes they veer off in an entirely different direction; sometimes they recreate the exact same look and just keep it consistent (which a LUT alone cannot do). Every project is different, but there's often pressure based on the nature of the project, the budget, and the schedule involved.

Again: the advantage of actually doing this work as a color correction and not a LUT is that it works in any color space, and this is increasingly important for projects that go to broadcast, to online, to theaters, and to HDR at multiple brightness levels. LUTs also don't work in ACES (for the most part), so that's another potential challenge for clients who demand to work in that format, like Netflix and several other studios.

 

[Michael Tiemann]

That was a really interesting video! Thanks, Marc!

 

[Björn Benckert]

My problem. My stuff never comes in even half evenly exposed. usually it´s all over the place, not only in terms of exposure. Bad ND´s and lots of different cameras and lenses involved... then you add sun / weather change, different light setups etc ontop of it. Then usually the subtile primers will simply not do the trick.

Taking it all into a common large colorspace might help. But usually there is tons of things that need to be color matched that really did not match in real. Some shots is greenscreen, other shots are filmed later when the sun is up and we moved and possibly got a big red wall behind camera that sadly acts like a big red bouncer etc, etc... The normal state of things is usually completely haywire. So then usually giant leaps is needed in grading to try to get a somewhat match in between cuts. Lot of times using rotoscope mattes to simply separate elements of the shots and counters stear the unevens.

What Im trying to say is, these fine tuning tools might be great for some applications and can maybe pinball something in somewhat the right direction, give confidence on set and most of all give the colorist and offline editor some proxies that somewhat look like the DP intended it to look. But when the actuall grading is done I find best practice to just use it as reference and then hit reset and start from scratch and let the colorist do his work.

This is maybe not the best example as it has a quite an destructive grade with an aim to age the material. But if looking at the top / what was actually shot. Water and color of most things sway like crazy between cameras, time of day, camera angles, etc. Possibly a bit drastic example but even when people shoot in controlled environments, same camera, good set of lenses etc, light tend to be adjusted to optimize for each setup and then pretty much all is a new give for each and every shot and you then need some one to dial in the right colors anyway.

https://vimeo.com/318436500

 

[Hrvoje Simic]

My problem. My stuff never comes in even half evenly exposed. usually it´s all over the place, not only in terms of exposure. Bad ND´s and lots of different cameras and lenses involved... then you add sun / weather change, different light setups etc ontop of it. Then usually the subtile primers will simply not do the trick.

Björn, Primers correct more complex things then ND and scene tints.
Light variable is the easier part to correct and needs to be corrected manually. Primers do the rest.
Primers are not a "grade". They optimize what goes in grade so the grading process could benefit in stability, precision and consistency.

Sensor and base CS offsets are fixed.
These tools are scene-independent. They work in every situation.


When light goes off native balance advantage increases because highly specific and complex signal offsets accumulation gets avoided through sensor-tailored image normalization.




Helium, Jason Han

Ungraded imagery, same white balance.

Overal tone is not all. Not even close.




Skin tonal separation, skin volume, skin definition, skin texture.

 

[Hans von Sonntag]

2 years ago, I switched from video to ACES (ACEScct) and don't look back (with camera footage). On IPP2 based Red colour science I found that using as IDT Alexa LogC gives me nicer colour to start with. In everything else I totally follow Mark's elaborations on LUTs.

Hans

 

[Björn Benckert]

Björn, Primers correct more complex things then ND and scene tints.
Light variable is the easier part to correct and needs to be corrected manually. Primers do the rest.
Primers are not a "grade". They optimize what goes in grade so the grading process could benefit in stability, precision and consistency.

Sensor and base CS offsets are fixed.
These tools are scene-independent. They work in every situation.


When light goes off native balance advantage increases because highly specific and complex signal offsets accumulation gets avoided through sensor-tailored image normalization.




Helium, Jason Han

Ungraded imagery, same white balance.

Overal tone is not all. Not even close.




Skin tonal separation, skin volume, skin definition, skin texture.

So IPP2 see the skin as slightly greenish in that light, with that lens and white balance. What if she had a too warm tone to start with or to blue? you clamp all skin color towards that color?

 

[rand thompson]

You could remove the green from the skin in the IPP2 example. But I think what Hrjove's primers do are, with my limited technical knowledge, knowing that the Red Helium's color science renders certain colors, greenish skin in this example, a certain way, the Primers bring those colors, Along with other important properties of the image into a more technically accurate representation of those properties in the beginning of the color grading process without having to do what I did below. And I believe that whatever the delta between the Image properties captured in camera and under the different lighting conditions, The primers would compensate for those conditions, bringing them back to a more accurate representation. Well that's at least my understanding of it.

 

[Hrvoje Simic]

You could remove the green from the skin in the IPP2 example. But I think what Hrjove's primers do are, with my limited technical knowledge, knowing that the Red Helium's color science renders certain colors, greenish skin in this example, a certain way, the Primers bring those colors, Along with other important properties of the image into a more technically accurate representation of those properties in the beginning of the color grading process without having to do what I did below.

And I believe that whatever the delta between the Image properties captured in camera and under the different lighting conditions, The primers would compensate for those conditions, bringing them back to a more accurate representation. Well that's at least my understanding of it.

1) Learn to write people's names properly.
If the upbringing failed in developing character virtues which entail respect towards others, manifested through motivation to at least write their names correctly, or dyslexia presents a continuous challenge with one word, copy/paste works.


2) You have no idea what you are talking about.


3) There is a longer post on the previous page which someone invested time in, in order to help in understanding.
Which you ignored and chose juvenile unappreciative smartassery instead.


4) It doesn't work that way and that grading you did on top of the image, motivated by malice and ignorance, does not stabilize the image.
You are just attempting to copy the look based on some of many manifested results.


5) This is not the first time you are trolling this forum with nonsense.

 

[Marc Wielage]

What Im trying to say is, these fine tuning tools might be great for some applications and can maybe pinball something in somewhat the right direction, give confidence on set and most of all give the colorist and offline editor some proxies that somewhat look like the DP intended it to look. But when the actuall grading is done I find best practice to just use it as reference and then hit reset and start from scratch and let the colorist do his work.

Yes, that's generally my experience as well. However... I have run into cases where a DP is completely locked into a custom LUT they created and they love that look. Rather than fight them, I'll go with the flow (a bit under protest, but whatever) and we'll use it just to keep them happy. As long as the final image looks good, all is well. I'm a fan of Patrick Inhofer's technique (from MixingLight.com) of putting a node in front of the LUT, then the LUT, then a node after the LUT, just so we can do a little tweaking if it doesn't quite react as expected.

But I also see people who have 100-200-300 LUTs available and they endlessly bop through looking at this one... then that one... then that other one... and they wind up spinning their wheels so long, it might take them 20 or 30 minutes to come up with a look. To me, that just wastes time. If I can sit down, look at the Raw, hit some Raw settings and dive in and come up with a look without a LUT in two minutes... we've saved time. Even better if I can give the client 5 or 6 different looks and let them choose how we should lean. And that applies even when the material is shot inconsistently, or badly, or shot on a different camera, or the shot was done on a different day with a different DP... I'm going to manually pound on the knobs until all of us like what we see. I will sweat blood and turn the knobs until my fingers turn blue if I can squeeze another 2% improvement out of a shot. A LUT can't do that. And (as I said before) so much of what I do involves lighting different parts of the scene differently, we go way beyond what an overall LUT is going to do. We're dropping in flags and a little kicker and maybe a key light over here and a shadow over there, gradually changing exposure over a long walk from one side of the set to another, adding a tiny bit of eyelight, and maybe helping an actor of "a certain age" look a little younger... this is what distinguishes detailed work from just ordinary "one size fits all grading."

We agonize over this stuff, even for small films that maybe no one will see... but when the filmmaker really cares, and when they come to us wanting help, we'll bend over backwards to do what we can to help them solve these problems. And that's basically the difference between a LUT and a colorist.