Grading and LUTs, etc.

[Guest 89e2bdf5797fbbdc17c2cc6da1413fa0]

Regarding looks.... Matt's signature quote from David Lynch reminds me of an interview I read of his, or maybe heard... can't remember. The quote may actually be from the interview. But, the gist of it was an answer to a question about high definition. David Lynch was saying that he didn't much care for it at the time and felt it was waaaay too much detail. So much detail that the magic and mystery of the dreamlike illusion were being destroyed. And, that he preferred the softer constrasty look of old 16mm films where the viewer had to fill in what was in the shadows with their own imagination. 

 

Or something like that. I think there's definitely something to that.

 

I think that's from an interview about INLAND EMPIRE, when he was talking about shooting it on consumer SD camcorders. I think he compared the low resolution and macroblocking (which he liked) to the softness of grainy 16mm.

[jasondhsd]

What is the difference between a lut and something like filmconvert?

[Sean Cunningham]

What is the difference between a lut and something like filmconvert?

 

Fundamentally not much.  What's different than most of the LUTs that get passed around is that their transformation is based on real film.  They do this by studying and sampling the results of control images passed through both digital cameras and these same control images photographed on various film stocks.  Calculating the difference between reference value and color lets you push one towards the other.

 

Of course it has limits but what you get as a result is an approximation of the result from doing a telecine of negative or scanned negative on a Datacine, where the resulting digital file is an inter-positive.  The quality and accuracy of your "scan" will be directly proportional to the quality of your digital footage, in terms of dynamic range, color gamut and (and here's an important part) its exposure.  

 

Some critics don't seem to understand the "inter" part and assume this is, is designed to be or should be your "grade".  It can be, the same way you can shoot film, send it to the lab and get a print back.  This won't be graded it will be at a baseline such that if you were to have shot the film properly exposed you get a proper, representational image back.  If you're on a budget and can't afford to pay for a colorist then there you go, that's your film.  Otherwise, based on seeing this baseline you can now decide if it needs modification to match surrounding footage and you can decide how and how much you would like to change the photography to achieve a different look for artistic effect, etc.  

 

Similarly, once you've applied Film Convert you can either decide you like the basic transform or you can then grade and modify further.  Some have claimed doing Film Convert somehow limits further modification but that's absolute nonsense.

[Guest Ebrahim Saadawi]

Similarly, once you've applied Film Convert you can either decide you like the basic transform or you can then grade and modify further. Some have claimed doing Film Convert somehow limits further modification but that's absolute nonsense.

I was always against LUTs/presets and always thought of them as a ridiculous approach to color grading. But what you're saying makes a lot of sense actually, I might purchase filmconvert and give it a try, there must be a reason many seem to love it!

[Axel]

The difference between a LUT and a Look in the strictest sense is that the purpose of the first is to convert a 100% defined A to a 100% known B. FC offers 'lotsa looks'.

 

LUTs as looks can be created in many applications, most CC suites (such as Apples Color, the AAE plugin Color Finesse and of course Resolve) can export LUTs. With the aid of something like 'Lut buddy', every host app can also export LUTs. These tiny files are freeware or shareware, or at least they should be. 

 

Beware to apply a look-LUT before actual color correction, because:

 

Adding a print LUT actually degrades your image. You’ll notice it removes a lot of colour contrast and hue variation. And the film effect is actually quite strong. So if you’re after a clean, slick look or your after neutral images it might be easier to achieve without using the LUT.

 

Cited from here. Google for more free lut files. Buy Filmconvert, because it comes with a GUI and the most-desired 'film grain'. And, as my link says (scroll down), it does give better results!

[Sean Cunningham]

I was always against LUTs/presets and always thought of them as a ridiculous approach to color grading. But what you're saying makes a lot of sense actually, I might purchase filmconvert and give it a try, there must be a reason many seem to love it!

 

You should really only do it though if you're going for that look.  If you aren't really unhappy with the inherent look you're getting from your camera then it's maybe an investment that would be better spent (not that it's that expensive) on a video card upgrade, if you don't have a compatible card for using something like Resolve, or some other tool.  I love what it does when I want that look but it's not hard to see example works out there shot on nearly any decent camera where I don't think the footage or film would have been improved any by pushing it through Film Convert.  It would have been different, but maybe not really better.

 

Please don't misinterpret my desire to keep things fair with regard to its purpose or value with any kind of blanket advocacy for it being used all the time on all things.

 

I've seen, for instance, lots of video shot using various Driftwood patches for the GH2 that have a great look as-is.  It's often hard to judge the stuff being shot on various BMD cameras because it's so often poorly graded (or worse, un-graded).  I've seen RAW 5D footage that I don't think needs FC at all to look film-like enough for me.

 

You can download a trial copy and you should play around with it before buying.  I also suggest getting the plug-in version, not the stand-alone which is tuned for speed and not precision.  If you don't want to spend the cash for their one price buys all you must be very careful about which product you buy because they're all licensed separately and licenses are not transferable between product or platform because they use a really kludgy style of local key.

[Guest 89e2bdf5797fbbdc17c2cc6da1413fa0]

This is becoming a really useful thread  :)

[Sean Cunningham]

 

Beware to apply a look-LUT before actual color correction, because:

 

Adding a print LUT actually degrades your image. You’ll notice it removes a lot of colour contrast and hue variation. And the film effect is actually quite strong. So if you’re after a clean, slick look or your after neutral images it might be easier to achieve without using the LUT.

 

 

This advice makes the assumption that you're doing things wrong, or at least, not as well as you should.  It's also making the assumption you're applying some bull-shitted "film curve" that applies a uniform transformation without respect for the digital file's origination.  It's like good advice for doing things wrong which I don't get if it's coming from that Australian colorist.  Resolve doesn't clamp or integer-ize intermediate results at the output of a modification node.  It makes no sense.

 

You will lose information if you read in 8-bit footage, apply a LUT and then save it back to an 8-bit file.  Don't do that.  That's bad.

 

You will lose information if you think because you shot 8-bit footage you can grade in an 8-bit project.  Don't do that.  That's bad. (I don't even like working 16bit any more)

 

If you are working in a floating point project. there is nothing lost after applying a LUT.  There is no removal of anything.  Nothing destructive is happening.  I can crush an image down to practically nothing and expand it back out to its original form and everything is still there.  That's true in After Effects so I would be really, really surprised if Resolve did something boneheaded like clamp or integer-ize all intermediate steps between nodes.

 

You should really be operating in a floating point project and saving both intermediate and final imagery to a high quality, un-compressed format with enough precision to not be destructive.

 

edit: he's also referring to "print LUT" which, while highly anachronistic and a dubious proposition at best in any case, is very different than what FC is.  A "print LUT", if you were going to do such a thing, would definitely be the last step, if you were really after an old telecine look (but if you work without having it on most of the time you are in for headache and heartache doing your grade, being happy with it and then applying something like this).  It's analogous to AE's View->Simulate Output->Kodak 5218 to Kodak 2383 (which would only really look right if you were working with scanned 5218 negative though they also have a "universal camera film" to 2383 as well).  

 

When you watch a movie on TV or BD or DVD or LD, ideally, you're not seeing the influence of a print stock.  Likewise with commercials shot on film.  They scan or otherwise transfer from negative.  If a movie has had a DI odds are no video representation you have or will ever see has any printing influence.  

 

Ideally you're always wanting to be looking at some representation of the IP and not something that's been "stepped on" because, yeah, you don't grade/time a film once you've printed it.  Maybe a "print LUT" would be useful if you were wanting to simulate the very specific look of old (like '80s and earlier) telecine (transfers from print rather than negative material) or something like Technicolor's ENR or other silver retention process applied to the printing of a film, creating an image that you would only ever see in a theater from select, expensive prints and not your typical release prints.  This is a look that you would not get seeing the same movie digitally projected or on a home video release.  

[Nikkor]

Actually there are a few things in Resolve that are destructive, I saw some videos about it when the ML raw happened.

[Axel]

 

This advice makes the assumption that you're doing things wrong, or at least, not as well as you should. 

 

Yes. Would the one who reads this thread and does best already please stand up? 

 

I know you are standing.

 

Similarly, once you've applied Film Convert you can either decide you like the basic transform or you can then grade and modify further.  Some have claimed doing Film Convert somehow limits further modification but that's absolute nonsense.

 

The proper way to deal with it seems to have to do with an understanding of the workflow, right? I don't think there is actually a danger with 8-bit processing, because all CC-apps can calculate in 32-bit, and most modern NLEs use it too (there were, however, filters and effects in legacy FCP and older versions of Premiere that were 8-bit, and one of them allegedly made everything 8-bit).

 

 

If you are working in a floating point project. there is nothing lost after applying a LUT.  There is no removal of anything.  Nothing destructive is happening.  I can crush an image down to practically nothing and expand it back out to its original form and everything is still there. 

 

The destruction isn't done by the software. It's done by us. I change parameters with every stage of the process, and from there I go on, in great parts in a WYSIWYG-fashion. Doesn't help much that everything remains reversible, there are just too many ingredients. The difference between me and a good colorist is a lot of botching around, trial and error and a growing uncertainness here and a goal-oriented, sensible order of operations there.

 

We want to be able to hit the nail on the head with any given look. So, understand, I'm not 'critizising' FC.

[Bioskop.Inc]

@BurnetRhoades : So basically you're saying that FC is v.good for getting digital Log footage out of its flat state & into a more celluloid looking environment. But you still need to or should i say, you still can then go away & CC/Grade to really fine tune everything to your taste.

 

So far with my BM Pocket footage i've been able to achieve everything I want in FCPX, you've just got to follow numerous correction steps (6 minimum) & build up/re-build the footage before it looks really good - none of that muddy footage for me.

 

So will FC enable me to get nicer results (will it really look like film) or is it best to just do it the long way a few times & make pre-sets.

[Sean Cunningham]

I cannot say for certain, in all cases, that FC is good for getting digital LOG footage out of its flat state.  I haven't experimented enough with LOG footage to know the extent of its capabilities in this regard or bothered to see what flat/log profiles they've created.  I have brought in some very LOG looking, still very flat, green, "typical" RED Epic footage into FC and it did what appeared to be a very good looking transformation of the footage.   It turned it into something instantly recognizable as "film like".

 

I do forsee an ACES workflow as being the more general purpose means to get better log->lin conversion than what is currently standard and I drew parallels to how that system works and what's going on under the hood of Film Convert.   Those six steps and that poking around to get a decent looking image from BMD cameras (or RED or any raw), either RAW or with that log curve baked into linear ProRes, is what the ACES workflow is designed to remedy, because it's the result of an incomplete, inefficient and error inducing workflow (no judgement on you, it's been that way for everyone).

 

But, back to FC, yeah, you can do your conversion and then consider this a new starting point to then actually grade your footage or you can just be happy with the primary conversion.  That choice is up to you and your desired look.  If you want your footage to looked DI'd and heavily modified then it would be best to do the conversion to FC within Resolve or AfterEffects and then do all the grading that you would normally do to any footage.  Since the whole DI process began with scanned film anyway, not digital camera footage, you are essentially turning your digital footage into some approximation of a scanned negative and now you're free to grade to the look you want like anything else.  All you're doing is altering your starting point.

 

Or, you just do the primary conversion and minor balancing for any exposure and color temp inconsistencies.  This would be analogous to what you see in film prior to DI.  Watch a movie from the '80s or the '70s or even all through the '90s when there weren't a lot of ways to modify the look of film, no such thing as secondary CC for feature film (though there was for commercials), just basic color timing and exposure before involving an optical house for some kind of effect.  There are thousands of films that are basically just lightly modified, processed film...basically everything before O Brother, Where Art Thou?  That's an approximation of what you have just running FC and not doing a lot of grading afterwards.  

 

There's nothing wrong with either way of using it.  I will say, it's kind of refreshing, going back to watch pre-DI films where folks weren't doing heavy handed modifications just because they could.  The opposite would be something like that new Liam Neesen film where he's an air marshall.  Hoo-ah, that duo-tone taken to absurdity looks like hammered shit, pardon my French.

 

Again though, download the software and play with it in demo mode for free.  It does give you film-like results like nothing else you can currently buy and better than any attempt that's come before it.  

 

One of the negatives I've read about it that I wholeheartedly agree with is that their grain is too strong.  It doesn't bother me that it comes up by default with 100% application.  The only thing that makes sense is that it would be either all on or all off.  Complaining about a program's defaults are silly to me.  That said, the scale of their grain is set based on selecting a gauge of film, from 8mm up to various 35mm formats and their 35mm options seem too strong.  

 

I believe their grain files are something like 4K but I don't think that they're doing an appropriate supersampling of the highest resolution file and transforming it to be appropriate for the frame size.  At the very least, it's all too big for 1080 footage.  It almost looks like they're doing a 1080 crop out of the middle of a 4K grain scan.  It would be more appropriate to scale down and filter.  Having worked with scanned 35mm film since 1993 which is not long after film scanning became common, I can tell you that grain is bordering on sub-pixel in a high quality 2K scan.  You might get something pixel sized in the blue record but the grains in FC are huge by comparison and, if nothing else, it just needs to be "kissed" in.  I'd suggest doing this after everything else since you can apply a second copy of FC with all of the color and film settings dialed down to pass-thru and just apply grain, or use some of the grain files that have been posted by other professionals for other folks to use.

[dhessel]

Just to be clear some luts can cause a loss of data. Some luts have flattened curves near the black and white ranges that case cause a loss of color information that is not recoverable. One example for colors on a 0 to 1 scale, black to white, a lut may expand the colors in the mid values pushing a value of 0.25 to 0 and any value from 0 to 0.25 also to zero.

 

Regardless of your working environment being floating point 32bit or not there will be no way to recover the crushed blacks in your footage. Now this is an extreme example but it is not that uncommon to have luts with flattened curves that can do this on either the black or white end of the spectrum.

 

To make this effect even more clear imagine a "black and white" lut that takes any color and makes it grayscale, once you have applied that lut there will be no way to get your colors back.

 

Also is you were to push some colors to super white and apply a lut you will most likely loose them and they will no longer be recoverable, again regardless of working environment. That is why it is recommended to use luts at the very end of a color grade.

[Sean Cunningham]

In such hyper-mundane cases you wouldn't apply something like Film Convert last because its result is based on a pre-grade profile.  If you're pushing everything so far out of whack that now values that once existed no longer exist the calibrated beginning profile now is meaningless.

[Axel]

So far with my BM Pocket footage i've been able to achieve everything I want in FCPX, you've just got to follow numerous correction steps (6 minimum) & build up/re-build the footage before it looks really good - none of that muddy footage for me.

 

This is the way to go. It's what you learn in every CC tutorial, you need to enhance the image step by step, in logic order, be it through a line of stacked corrections (FCP X), rooms and multiple secondary CCs (Color) or a node tree (Resolve). Enhancing means knowing what you want to achieve before and get there by successively 'destroying' (in the sense a sculptor takes away material, not as degrading) and /or adding and changing values.

 

And why not take a neutralized (optimized for exposure and color fidelity) primary CC and apply an FC stock emulation? This could add a very subtle look.

 

The most prestigious task is not finding a pleasing look (it may be for a music video), it is getting the style right. No haphazard experiments with flavours one has no idea from which elements they are composed.

 

For a feature film, the director and the production designer often start with moodboards. Those are collages of photographs, sketches and bits of fabrics that help define how the film should feel. This predetermines a lot of all the things (including costumes, sets and lighting) contributing to the look, let's rather say style, desired for the final film. 

 

Ideally, imo, I am aware what I want before I start grading. Obviously I was wrong when I accused FC of muddying the water. But it can, if you are not aware of what it's all about.

[Bioskop.Inc]

Well i tried the free version (doesn't have too many pre-sets, are there more in the paid version?) & it is nice, especially the shadow/highlight roll off (the grain is nasty, really nasty - yuck!).

However, when i compared it to a few tests i'd been doing I found that i wasn't really that far off after all - mine were slightly more contrasty, with a bit less saturation in the mids/highlights.

So i think it might be best to save some money (those Speedboosters aren't cheap) & learn something in the process.

[Sean Cunningham]

Well i tried the free version (doesn't have too many pre-sets, are there more in the paid version?) & it is nice, especially the shadow/highlight roll off (the grain is nasty, really nasty - yuck!).

...

 

 

Do you mean for film or for cameras?  There are a bunch of camera profiles in their download section so if the trial didn't have your's check there.  As for the film presets, yeah there aren't that many but there really wouldn't be.  You have a few Kodak stocks, some Fuji stocks and then there are some stills stocks.  There are a couple of Kodak stocks that I wish they would do that are contemporary and still being used but they haven't profiled it yet.  Most of what is there I've ignored unless I was just seeing what something more "oddball" might look like.  

 

It's virtually impossible for them to profile stocks that are now discontinued.  And since it isn't an Instagram like product they've kept it to fairly practical choices.  There weren't that many choices ten years ago and there are fewer today if you were going to shoot a film on film.  You have a couple choices from Kodak and a couple choices from Fuji.  Looks like Agfa just offers a couple, a B&W neg and a color print stock.   They have, however, needed to have Kodak Vision3 5219 500T since the beginning as it's one of the most used stocks today.  

 

Fresno Bob made one of the best demos I've seen of the available stocks they've profiled:

 

 

 

Here's a typical progression for how I work, from straight GH2 (in this case it was Flowmotion patch), Film Convert (Fuji Pro 160s), ColorGHear grade (enhance saturation, play up the neon in the scene), final recipe for micro-contrast to bring out skin and tiny black details:

 

[Sean Cunningham]

This is the way to go. It's what you learn in every CC tutorial, you need to enhance the image step by step, in logic order, be it through a line of stacked corrections (FCP X), rooms and multiple secondary CCs (Color) or a node tree (Resolve). Enhancing means knowing what you want to achieve before and get there by successively 'destroying' (in the sense a sculptor takes away material, not as degrading) and /or adding and changing values.

...

 

Except that most of that work isn't actually enhancing the footage, it's getting to a reasonable place to start.  You can't intelligently grade your footage until you see it for what it is.  That's not what you really start out with most of the time when shooting raw or if you've shot linear with a baked in LOG curve.  

 

If I was a client and I'd shot film and I got rushes back that were different every time or, worse, I was at a lab timing my film and having to sit there and watch the colorist poke around and try this or that to get my footage to simply look as filmed and as good as I would expect from rushes with house lights I would take my film and go someplace else.  But somehow this is how folks are expected to work with the digital equivalent to a film negative.  It's awful.

 

Once you're really grading, I wholeheartedly agree, methodical, logical order.  But it's easy enough to see in upload after upload that most folks aren't starting their grade from a good place before trying to get to their look.  Just look at Shane Hurlbut's BMD tests.  They're no better looking than Joe Six Pack bought himself a BMCC's uploads to Youtube.

 

I've got this fellow's entire short shot on an Epic sitting on a drive.  I look at the footage as-is in After Effects it looks just as bad as it does in Redcine-X.  Just awful.  Unusable for my purposes.  So I'm having to linearize it all to a meaningful place to work that's appropriate for display on an sRGB monitor.  That extra work is eating into time that might otherwise be spent doing my actual job on this fellow's film and it's going to be the same waste when he goes to do his final grade in Chicago because the colorist is dealing with it like everyone has always had to deal with it.

 

At least most of these cameras seem to have decent monitoring now, even if you're going to have to work to get it back to looking like what you remember from set once it's shot.  One of the most said phrases by Brian Singer on the set of Superman Returns, the first motion picture shot on the first model Sony Genesis, "it's not going to look like that, right?"  We have come at least that far.

[Bioskop.Inc]

It's virtually impossible for them to profile stocks that are now discontinued.  And since it isn't an Instagram like product they've kept it to fairly practical choices.  There weren't that many choices ten years ago and there are fewer today if you were going to shoot a film on film.  You have a couple choices from Kodak and a couple choices from Fuji.  Looks like Agfa just offers a couple, a B&W neg and a color print stock.   They have, however, needed to have Kodak Vision3 5219 500T since the beginning as it's one of the most used stocks today.  

Ok, i thought so & what they do have is v.nice, as a starting point or a finished look. As i've said the highlight/shadow roll off is lovely & almost impossible to replicate - I can get one close but then the other suffers as a consequence.

 

 

Once you're really grading, I wholeheartedly agree, methodical, logical order.  But it's easy enough to see in upload after upload that most folks aren't starting their grade from a good place before trying to get to their look.  Just look at Shane Hurlbut's BMD tests.  They're no better looking than Joe Six Pack bought himself a BMCC's uploads to Youtube.

 

So by good starting point, you mean get the exposure & contrast right (for BM cameras, put the life back into the flat film/log files that come out of the camera), then you go for the grade - that's how i've been doing it recently.

[Sean Cunningham]

 

So by good starting point, you mean get the exposure & contrast right (for BM cameras, put the life back into the flat film/log files that come out of the camera), then you go for the grade - that's how i've been doing it recently.

 

Exactly.  You're doing work before doing your work, I'm assuming pretty often.  Ideally you should never have to look at flat, log footage and begin (it's not meant for human consumption, not at all) or, on your own, push it around until it's a correct representation of what you shot for your display.  The notion some folks have that when you shoot raw the image is totally up to you and arbitrary and it's in there and you just have to find it is nonsense.  They've been sold a bill of goods and cameras before software was really ready to handle the footage properly.

 

The process that's been standard up to now is incomplete and not fully realized.  That will change though.  It takes getting all the camera manufacturers on board and the software companies moving towards a standard though.  And, unfortunately, a lot of re-education of some working professionals and enthusiasts.  Until then we have to make the best with the tools we have and ride out the one-step-forward-two-steps-back reality that comes with certain advancements where software is often left behind by hardware.

 

What's going on now reminds me of the early days of scanning film.  You'd get your 10bit log .CIN files that needed to be linearized so that it was actually useful and looked like a real image.  You'd get a match clip from surrounding footage and then someone would have to painstakingly, going back and forth from monitor to loupe and light table, re-create something meaningful out of the scan.  In a facility of a hundred or more artists you might have one, or two or maybe three people who could adequately perform this task.

 

Flash-forward ten years and the process had only gotten a little better but you still had a similar process happening that created a "show LUT" that might vary even if show-to-show scans were being made of the same film stock.  It was terrible.  Labs don't have to re-learn how to process the same film over and over again but the process would be changed project to project with digital files because it was never truly right or ideal.  Thankfully those days are coming to a close and when you load a CinemaDNG or R3D or .CIN your software will already know how to display it properly and you can concentrate then on how you want it different from how it was shot.