Grade School: Filmic Grading Principles

okay everybody hopefully i'm live and you guys can see me this is my first time doing a youtube live so i guess i'll wait for a couple folks to populate and uh if someone wants to chime in on the chat and tell me you can see and hear me that'd be awesome okay i see belinsky there the vdap nice to see you guys patrick's in the house all right someone let me know when you get a chance can you can you see and hear me okay everyone should be able to see my screen and my beautiful face okay thanks rob awesome all right guys i'm gonna get into some stuff you guys can feel free to chime in with any questions you've got it's open territory we can talk about anything that we've done uh in recent videos other color grading questions you've got and i'm going to try to get to everybody as best i can for the moment i actually want to start with a really good question that i got from philippe earlier this week and i hope felipe will be able to join us this morning but felipe asked that it seems obvious that narrative fiction does indeed have high contrast right so he says the the contrast is high and the colors are dense so where does the idea of film and images for narrative being low contrast come from everywhere on youtube there are camera guys who think they're experts telling people that you need lower contrast to achieve a desaturate and to achieve a desaturated look etc which of course when you do capture footage like this on 422 8-bit cam is a disaster to recover in post and his question is any idea why most people think uh who who think they're interested in filmmaking think the opposite of what is actually true so really great question and uh this is one that's near and dear to me my heart that i've talked and thought about for a lot over the last couple of years i'm going to answer it by dropping one of my favorite quotes it's an einstein quote that you guys might have heard before where he talks about the idea that we want to make things as simple as possible and no simpler right so i think that's a bit of an applicable idea when it comes to talking about contrast and i believe felipe's comment was actually on my create better contrast series where we're going into a bit more detail on okay well we've got contrast and it actually turns out that we can think of almost everything that we do when we're grading as some kind of contrast adjustment we might be adjusting the distance between different tonal regions or between different uh colors in our frame or between different spatial regions of our frame but a lot of it fundamentally boils down to contrast almost everything that we do so it really pays to start to get a little bit more granular granular and a little more specific in terms of the way we think about contrast so i think that's what's really the issue here when we talk about this sort of wave that i very much see too i think felipe asks a great question i see we've kind of got this like low con like like aesthetic that has really come into vogue over the last 10 years and some of it comes probably from the so-called log look where there's actually a bit of a knowledge gap where people don't understand that when i'm looking at a piece of footage in its log form like on a shot like this that that's not a valid aesthetic starting point that's what the camera captured and that's not the same thing as what our display can reproduce which we've talked about in other videos but all of that said my short answer to that question would be that i think that what people are looking for when they talk about wanting to control or reduce contrast they're actually not looking for less contrast they're looking for more depth because if we look at actually a shot like this is a pretty good example i'm going to go to my timeline level and i'm going to disable my output node and i'm going to assume that uh hopefully most of you guys who are here today have been somewhat following along with the videos if not i'll do my best to kind of get you up to speed on the way i roll i'm just going to turn off my technical output transform here which is what's putting me in direct 709 so without that the rest of these nodes are all uh empty at the moment so i've just got a stock area image here and let's just uh you know kind of break down this contrast thing for a second if i grab for my contrast knob and just crank a bunch of contrast and go to what i would consider a high contrast look let's grab a still of that and i'm going to turn that off and now let's turn my technical transform back on and now i'm going to turn on one of my trusty film prints and let's wipe between these so maybe i added a bit more contrast uh in my first version probably not honestly it's about the same like overall level of contrast if we think of contrast as a single variable single number but the results are very very different right so this is sort of my visual answer to that question of it's not really about high or low contrast it's equally as much about what kind of contrast where are you introducing it is it rolling is it linear is it happening in the log state is it happening on the gamma state of the image all these details that end up amounting to good or bad contrast so hopefully that kind of fills in some of the the blanks there for uh philippe's question but i kind of wanted to start by touching on that because it's a really good question and actually i wanted to show you guys one more thing so i'm going to turn off my output transform and let's do a bit of a film print contrast curve worship for a second because one of the reasons that film and narrative fiction gets away with high contrast and why we think it looks so darn good is because a lot of that stuff is feeding off of legacy film print color curves or contrast curves rather so if i look at i'm going to turn this image into a ramp and if i look at my waveform down here when i turn this print stock off so i've got a linear ramp right now and then turn it back on you can see just how particular this contrast is it's got a big swooping roll in the bottom it's lifting uh the the uh overall like lift of my image a little bit so the blacks aren't actually reaching pure zero and i've got this big gentle massive shoulder on the top so the net of it as we saw on this image is some very strong contrast but it's a very particular kind of contrast and generally speaking that kind of rolling or s-curve contrast is the kind of contrast that we are going to find visually palatable and when we do a more linear form of contrast or as another example if we were working on this image say in its gamma state after it has been processed from log and into gamma and i were to go in and kind of work on the same experiments even with like a normal amount of baseline contrast here i'm just not going to get nearly as good of results they're going to feel more quote unquote video e because i'm not manipulating contrast at the right point in the processing chain and i'm also not introducing the right kind of contrast into the image so more than you ever wanted to know about my thoughts on uh you know the the sort of high con versus low con debate that uh happens so much and the fact that sort of in a a broad wave we've tended to move toward the lokon thing a little bit i would say in terms of filmmaking culture and i think that's really born out of the the desire to get more depth and not knowing how to preserve that depth while adding strong contrast it's kind of a a knowledge gap as opposed to an absolute aesthetic preference all right i'm gonna keep going here i can talk about this stuff all day you guys can absolutely feel free to chime in at any point with your questions uh so let me just keep going through uh some of the ones that i'd kind of flagged okay so this is a great one um was everyone watching uh the the most recent aces series everybody get a chance to check that out if you haven't it's i think very much worth watching not only as an introduction to aces but as an introduction to color management and uh scene referred grading in general scene referred meaning that we are honoring and uh taking into account the conditions of the image when it was shot as opposed to working in what we would call a display referred model which uh to use you know like let's just flip to a fresh image just uh for the sake of mixing things up so scene referred would be saying okay this is area log c and i need to map this into rec 709 those are two different display or excuse me two different color spaces with two different um sets of primaries tone curves etc so in a color managed or a scene referred model i would say all right i'm going to take all those variables into account and i'm going to produce a technical transform to do my sort of baseline processing or what we might consider our lab work so if i turn on this output node here and go full screen you guys can see what i mean this may not be creatively all that we want out of this beautiful uh test image from arie but it is normalized it's no longer like completely flat and devoid of contrast so this would be a scene referred model if i set up this technical output transform and i work underneath it which is uh the sort of foundational aspect of working in asus it is a scene referred color managed workflow the counterpoint to that would be a display referred workflow where if i turn this off and this is the way that i worked when i first started grading and nobody taught me any better and in fact the colorists that i learned from early on in my career all worked in this way where you pull in the image and you're looking at it and you go huh looks like it doesn't have much contrast i better add some in and you add some contrast into your image and then you probably go huh there's not much saturation in there i better add some of that in there and you're making all of your decisions purely off of what your monitor is telling you let's see let me see what uh patrick's uh note is there yeah so the full screen is not showing up when i when i pop up to it huh huh that's annoying i'll have to look into it it's showing on my monitor correctly uh let me just double check the stream here and see if i can troubleshoot it that way ah that's totally right okay you know what i'll do for now i'm just going to do the it's better than nothing i'll do this kind of slightly larger view as opposed to going full screen for you guys thanks patrick so where was i so back on the display referred model we're making all of our decisions based off of what our display tells us and there's a couple problems with this approach first of all our decisions are only ever going to be as accurate as a display that we're working on and we all know that no display is perfectly accurate secondly in a world like we live in in 2021 where i i can't really tell you guys the last project that i color graded where there was only one deliverable and there was only interest in one deliverable we need to be able to master four different display color spaces without having to subjectively guess and start over every time and when i work in this display referred model like i'm doing right now all of my grading decisions are only valid and only makes sense for this exact display that i'm on the moment i switch over to a different display things are no longer valid and i have to use my eyes to try to get a visual match between what i see on display number one and what i'm trying to get out of display number two so a couple of my beefs with the display referred workflow but reason i just kind of took a fresh lap through that is that's really the power of working in something like aces to be honest with you i don't care about aces like i think it's cool and i i definitely have used it and will use it on particular projects but it's a really good sort of diving board a good introduction sort of jumping off point for working in this color managed scene referred model which is much much better for a variety of reasons that i go into in that aces series so check it out if you haven't already but one of the really good questions that i got about aces was is there any advantage this is for from uh teomir i hope i'm pronouncing that correctly he asks uh is there any advantage to use aces if your footage is all in rec 709 and by advantage i mean greater quality dynamic range lower noise richer colors or something like that i tried it and all the controls luts and other stuff are behaving very differently which drives me crazy i hear you it feels different right it says but i'm wondering is there a point for me to keep trying and i thought this was a really great question because i feel like the value proposition is a little simpler when we're working with log color spaces with big uh uh you know like like uh higher quality image sources but what if we're grading like in rec seven or nine what if it's a bunch of iphone footage is there any point to working in a big fancy color managed workflow there is a judgment call to be made there because there is a point for me personally past which i'm not going to bother color managing stuff like if you're handing me footage that is so fragile and delicate like something that you shot in low light on your iphone from five years ago to make an extreme example probably not going to color manage that because it's so delicate and brittle as it is that i'm going to take as light as possible a touch as i can on that material and just the round trip of exploding it out into a log working space and then putting it back down into a display space has the potential to introduce problems and to sort of damage the image even further than it already is so there's a sort of edge there past which i would say you know what i'm just going to kind of feel my way through this and use more of a displayed approach like i just was talking about a moment ago however that's at the extreme end of a spectrum that i generally plot somewhere along as opposed to at the very end of it and the question of is there an advantage my answer would be absolutely a couple of those advantages would be number one if i have a show or a film where it's mixed source right where maybe some of the material is rec 709 but some of the material is area log c or canon c log or whatever the case may be if i go to the trouble of actually input mapping and continuing to work in that big asus color space i'm going to be able to unify all of my creative decisions i'm going to be able to paste looks from one source to another without having to take those sources into account after that initial input mapping but the even bigger question of okay what if it's all rec seven or nine what should we do there what's the the best approach there again i would say unless the footage is so fragile that it's not even going to hold up to that round trip which you can always audition i still like to do it for a couple of reasons the the fundamental one though is uh this concept that i talked about uh again in the asus series and you guys have heard me talk about a lot which is that i want to work photographically with my images and i can really only do that when i'm in a log state on my image when that image is in a gamma state it's just not possible so actually i want to show you guys here's a quick kind of demonstration of that idea let's go to that this image is actually fine so i'm going to turn good my output transform is back on and i'm going to show you just some exposure adjustment here now i'm using uh one of my colloid uh tools to do this which is doing like a true uh you know like photometrically accurate exposure adjustment it's really really close to an offset or even better using your global exposure tool down here in your high dynamic range this is a tool that just is a bit more accurate along those lines but i want to show it to you guys because when i'm working on my image in its original log state i can trim stop or i can open things up and short of the actual limitations of the sensor which captured this image i'm getting a very iris-like response with this knob and i have quite a bit of dynamic range to play with before i'm like oh i just can't go down any further because my highlights are too far clipped and it's looking weird or on the bottom it's like oh i can't go up any further because the shadows are too noisy i've got quite a bit of dynamic range before i hit any of those extremes when i'm working in this model now let me show you guys something i actually this is a tool like i said that i've built and i've been playing just for the sake of sort of proving this concept out with doing it in a slightly different way so let me see if i can find give me one second here i'm going to find kind of a beta version of this tool where i was playing with okay well what if i did deploy it because all it's doing is allowing me to manipulate exposure by linearizing whatever curve i'm in so i set up a preset for like a gamma 2 force like what if i'm in rec 709 and i want to re-linearize it and manipulate it in the exact same way let's look at what happens when i do that i'm getting pinned in the top there so like my my highlights instead of uh preserving and then ultimately coming down are turning into this kind of mushy gray and then if i try to go up you can see the response is just not anywhere near what i was getting before even though i'm applying the exact same math to this image and even taking into account the different tone curve of the image it just doesn't work when you're in a gamma space where all of your dynamic range or the majority of your dynamic range has already been clipped off so to go back to the original question when we input map our rec 709 material into a working space like asus cct we're obviously not improving the image it's not going to get any better or magically have more dynamic range stored within it but we're ensuring that every inch of dynamic range that is in there we have access to it and we can manipulate it photographically as opposed to what i call graphically which is the kind of way i was taught when i began working as a colorist of playing with my lift and my gamma and gain and fussing and finessing and a lot of that stuff we turn all these different knobs and most of the time it's really just in the interest of getting a perceptually linear response so if that makes sense so the the response that i'm actually looking to get most of the time or at least me personally when i first started grade and i was playing with lift gamma and gain all i was trying to get the image to do was this just trim exposure or give me more exposure and to make that work when you're in a gamma space you really do have to go in and noodle with all three possibly even four of these settings and subjectively find the right depth for them to work at so very long answer to the question of is there still an advantage to working in aces when all of your source or the majority of your source is rec 709 my answer would be absolutely with the caveat that if it happens to be very very delicate footage you want to just test that out and look at the pipeline and make sure that you're not introducing artifacting or other problems into your image just via the round trip of going into asus cct and then back out to rec709 so i hope that kind of makes sense all right let's see how we're doing in the chat here oh that's a nice question uh from uh sorry my screen is too small uh sue habe i hope i'm pronouncing that correctly can you explain between a no look look versus straight out of camera so let's talk about that this is something that i often do with clients because i i feel like it's really really helpful to just kind of calibrate our vision and our expectations about like where do looks come in where does what do we mean when we say raw or straight out of camera or processed let's kind of chop that stuff up a little bit so i'm going to go back to my output transform over here and uh you know for those of you guys who may have just joined or who haven't checked out our asus series all this out compound node is doing is taking me from my camera color space what the camera saw and into my display color space meaning what my display can reproduce those two are not the same thing they're different and we the but the good news is we can account for that difference mathematically as opposed to just having to visually subjectively guess at it based on what our monitor is telling us so that's my kind of final output mapping of my color management that i'm working in here but let's talk about sort of like the color space journey or the the look journey if you will uh the entire journey of an image from capture to reproduction so right now we are looking at exactly the image which was encoded onto the hard drive from the sensor of this area alexa camera when this was shot i'm not doing anything at all right now to this image so let's talk about kind of how the process goes forward from here generally when you are shooting this image through an alexa you are going to be monitoring it through some kind of output transform most commonly aries alexa the k1 s1 lut that they provide with the camera it's a pretty good lut it's a very very neutral transform it's flattering but it's really not doing much with the image so i would call that sort of uh the straight out of camera in the sense that like straight out of camera is what we're looking at right now is technically straight out of camera because we're not doing anything at all to it but straight out of camera and properly prepared in a technically sound fashion for our display that's this right here and even that right there this subject is so interesting because the more you explore it the more you learn that there is to learn but even that simple question of how do we transform from our camera color space to our display color space it gets interesting really fast because my display can't reproduce nearly all the color and all the dynamic range that is contained in this source image so there's some judgment calls that have to be made like all right which pixels are important which colors are important which values are important because saying that i'm going to convert area log c and direct 709 is roughly the equivalent of saying hey you see that 10 pounds of dirt over there put it into this five pound sack well i i can put five pounds of that dirt into this five pound sack and i can be judicious about which five pounds i put into this five pound sack but i cannot put ten pounds of dirt into a five pound sack so even the question of an output transform is uh a bit tricky because there's different ways to get from that big high or a wide color gamut wide dynamic range down into the narrower dynamic range and color gamut of a rec709 display but this is one technically sound solution to that math problem of how do we transform from camera color space to display color space so this is what i would call a sorry i keep going full screen and i keep forgetting you guys can't see that that so i'm going to try to remember to just do this kind of scene view thing here so this is what i would call a no look look to answer the question there's really very little opinionated manipulation going on here there's a little bit like i said in terms of the judgment calls being made with the particular output transform that i'm using but the goal of that transform is to create a reasonably pleasing but aesthetically neutral transformation from camera to display so that's kind of a no look and then from here is really where my work would typically begin when i'm working in color i could work uh in in and just apply things by hand as you guys have seen me do generally what i would want to do here is do my global contrast curve next which i could do with something like my custom curves and just kind of go by feel one thing that i try to pay attention to here is to make sure that i'm preserving middle gray so that i'm not inadvertently pulling or pushing the exposure of my image just by trying to modulate the contrast or if i didn't want to do a global contrast curve by hand this is generally what i use a film print look for and this is one that i've created and prepared for my workflow that works inside of this area color space and as you guys can see we all have our own aesthetic preferences this lut which i've spent many years refining and working on and is one of a few that i i have at the ready for most of the work that i do you know barring client preferences and direction just going off of what i dig like this looks awesome to me and it's most of what i'm looking for out of my image just out of the uh out of the box with one set up in terms of like an overall look so i that's a long-winded answer but i hope that kind of answers your question about like neutral versus unprocessed versus actually applying a look let's see what else everybody has to say here okay i'm going to grab zeke's question right here because i see it right at the end zeke says using a film print lut at the end of the pipeline given the stock film luts in resolve in a scene referred grading environment could you explain the difference of using rec 709 luts and the log luts so let's have a look it's been quite a while since i looked at these tools so zeke i might have to have you clarify that for me so i have the rec 709 and the dci p3 luts which are about the target display they're actually both expecting a log source they're expecting like a cineon input so that's once again a scene referred uh environment that is uh based off of a uh a film scan um so that's that that's kind of the idea there's these are actually set up to be a one-stop shop scene referred color management pipeline with technical transform and with a look all baked into one single lut obviously that's a very simple uh approach and if you like the look it can be a very uh efficient approach there are some downsides to having all of your technical transform and all of your look cooked down into uh one lut uh so the the these your mileage is gonna vary uh in my opinion with these stock film looks i i used them and and and tried to massage them as well as i could for many years but they're they are they have limited utility in my experience all right let's see what else we got here i gotta get caught up you guys have lots of questions uh easy one do you keep loom mix value to zero no stock answer there uh generally i will leave it where it's at in the default on resolve it just all depends on what i'm doing but i also should point out that a lot of the contrast manipulations that i'm doing at the outset of a grade are actually not even coming from the primaries bar you guys saw me talk about exposure a minute ago that is i i try to get as far as i can with exposure before i even improperly manipulating contrast and then beyond that for contrast i'm generally using my contrast plus tool which i built and if you guys look at this here it actually effectively has a loom mix knob built into it so you know this would be the equivalent if i pump some contrast in this would be the equivalent of the low mix set to zero and this would be the equivalent of the lumix set to uh one so that i kind of i i see it as a creative tool that i i work with uh you know and and move around depending on what i'm wanting to do with my image let's check out luke's question is there a difference between putting the output transform on the timeline node versus putting at the end of your node tree great question no absolutely not you can put it at the end of uh the the node tree at the clip level and that's totally fine and that's a totally valid reason for wanting to when you want to have the flexibility to put a node after that transform in some instances or maybe you even want to leave it off for certain clips so no the the order of operations is uh you know like the end of the clip section of the node graph ends and then the timeline node graph picks up right after that point so the sequencing is still totally valid there all right let's see what else okay how about a qualifier i looked at the lessons on mixing light and there is a slightly different workflow where there is space in the log for keying how true is this working process uh talking about lessons about aces okay you know that's a good question i i i'm not sure i can specifically answer actually i'm live um i'm not sure i can specifically answer about um the the tutorials that you're asking about but in terms of the qualifier that kind of cues off another question that i got this week that i thought was a great one where you know like let me kind of reset all my business here for a second and go back to my stock setup of this image here and let's just talk about pulling a qualifier there's a what i would call a misconception pre-resolve 17 that qualifiers are not easy to pull or more difficult to pull in a log space i haven't found that generally to be true and supposedly i have to say i still haven't been able to 100 verify this but if you read the documentation in resolve for resolve 17 supposedly the qualifier is now color space aware meaning that it makes literally no difference where uh you pull your qualifier in your node tree so it should simply be a matter of you know using your highlight mode to dial things in a little bit by hand and i do want to just say one more thing about the qualifier while we're on the subject because i can't resist i want to point out to you guys i'm just going to give you a hard and fast rule that is going to change your qualifier game overnight okay you see these three parameters that i have here my hsl qualifier never never use all three just don't do it you're gonna find if you actually go through and look at like let's just do a quick example here let's say oh i want to pull a skin qualifier which i wouldn't do there's better ways to target skin but let's just say that i do want to do that i'm going to go in and i'm going to pull this and let's go to my highlight mode and i'm going to go full screen or close to and zoom in now let's look at what each of these parameters is doing i can flip them on and off with these little switches here and what you're going to find almost any time you pull a qualifier is the majority of the qualification is coming from no more than two of your your uh parameters in this case saturation and luminance are kind of doing it for me i guess it's sort of including my maybe it's more like hue and luminance but you guys can see what i mean you rarely need all three and the reason i say never do all three is it's almost impossible whether you're in gamma whether you've got the most beautiful footage in the world or whether your footage is horrific it's almost impossible to pull a really really clean really really easily pastable key when you've got all three parameters enabled and that's one of the things that i try to keep in mind when i'm pulling qualifiers i think of qualifiers as like i don't know if i'm like a bus driver and there's loud kids on the bus that i have to keep an eye on qualifiers are the loud kids on the bus like i have to keep an eye on them for the rest of the trip to wherever we're going i gotta keep coming back and checking on and make sure hey did you put some noise in there when i wasn't looking hey when i pasted you from shot three to shot four did you add some noise in there or did your key get chattery they're very expensive to babysit so i try to avoid them in general and when i do use them i try to use them as broadly as possible so that i'm not introducing those kinds of problems into my image so more than you asked for there but hopefully that answers your question and then some okay let's see what else is there any ways to convert rec 709 to p3 color space absolutely easy easy easy to very easy to do i've done a number of uh rec 709 masters for films before that you know we decided hey we we want to master it in 709 for you know whatever maybe the primary deliverable is tv or maybe that's just where everybody's comfortable maybe that's the reference display that we have and then we want to create a uh dcp uh you just do a technical color space uh conversion from rec 709 to p3 it's quite quite quite simple to do in a technically sound way it's not the exact same thing is actually monitoring and mastering in p3 but pretty darn close let's see i hear that i see this question about some tips on getting the mirandi look with pastel colors uh i'm actually not familiar with the morandi look but certainly a look with pastel colors absolutely would depend a lot on the original footage and art direction it's funny i'm working on a project this week that between us girls is for it's one of the first i've had in a while where a lot of what i try to do with like harmonizing color palettes and sort of like articulating like where do each of those colors live and how saturated should they be how dense or how bright a lot of that work was actually done really really skillfully by the production designer and the art director and it's a game changer like the extra distance that i can go for this client because they were already so very like mindful and careful about the way they constructed that stuff is night and day so there's absolutely it's impossible to overstate the importance of gathering a really really good image that is as much in line with the creative intent of the author as possible it just means i can do more when we get to that stage okay i'm gonna go to another question that i'm kind of burning to answer that i got uh earlier this week from clayton i don't know if clayton's with us today but uh give a shout if you are clayton this was an awesome question he says hey colin i discovered this channel yesterday and i think you're wonderfully articulate oh shucks i'm going to start blushing i'll skip that part he says i'd like to ask you a question i've seen a lot of colorists including yourself choose to use a log gamma as their working space could you elaborate why this choice is often made and he goes on to describe the fact that coming from a vfx and compositing background he understands correctly that working in a linear tone curve as opposed to a gamma or a log tone curve is the way that light out in the real world works before it hits our eyes before it hits camera sensors before displays come into the mix linear is the way that things work out in the real world and so clayton's fundamental question was why haven't colorists decided as vfx artists by and large have decided to use a linear tone curve as their sort of stock working uh tone curve if you will and he mentions log gamma in here and i'll i'll just clarify you guys who've been hanging out with me for a while know that i i like to be specific about my use of the word gamma to refer to a particular kind of tone curve but when i'm not referring to that specific kind of tone curve i will call it a tone curve and not a gamma curve so just in case that confused anybody in the question i always found it a little confusing when i was first learning is like wait are we talking about a curve or a specific type of curve so when i say gamma i mean a specific type of curve when i mean a when i'm just describing a curve i will generally refer to that as a tone curve so let's chop up this whole linear log gamma thing so there's a couple things going on there um and i'm really glad clayton asked the question because the short answer is we should be using linear light as our baseline as our default when we are working in with manipulating our images so just for an easy example of that i'm going to kill this qualifier i'm going to turn my my look node back on just because i like it and we already looked at this exposure tool here earlier but it's worth noting that that is exactly what this tool is doing it's taking the tone curve that i tell it that i'm in it needs to know what tone curve i'm in which is log c and it's linearizing that tone curve and applying linear gain because that is the textbook photometric definition of exposure of adding or reducing it so that's just one example of the best way to get exposure in or out of your image with the cleanest results is going to be to apply linear gain taking your source gamma into account so that's one example the other one uh right here downstream but this is one of my other go-to tools called points works the exact same way it works on the model of printer lights or printer points but it works by linearizing the input tone curve and applying global uh gain up or down the only difference is it's doing it per channel in addition to doing it globally and it's also doing it in the traditional unit of printer points which is uh helpful for uh you know anyone who's kind of coming from that world or or finds it to be a meaningful uh frame of reference uh when you're making small adjustments like hey let's add like you know a quarter point of green or magenta or whatever it can sometimes be helpful shorthand so that's kind of like the the broad answer is we should be working in linear more than we are the next thing the the question that comes up is like well why is it that we're not and it's a big question but i'm going to answer it as succinctly as i can and say that it's really a question of these two warring disciplines that make up the the modern color grading culture if you want to call it that there's sort of two distinct headwaters to that culture that you guys have probably heard me talk about before we have color timing which has its origins in narrative filmmaking and photochemical lab processing and then we have color correction which has its origins in broadcast video and they're very very different and we can see the tension in between these two traditions kind of being forcibly melded together over time even in the interface of resolve like the fact that when i open up resolve and i see in my primaries i see lift gamma and gain those are color correction tools those are really not tools of color timing color timers in fact only had one tool available to them and it was this points tool that i just showed you guys it's global printer points of i want to add or pull red green or blue or globally increase or decrease my exposure that's all that the color timer had available to him or her one sec i got so excited i dropped my pen um so we are dealing with this sort of like uncomfortable bedfellows thing with color correction and color timing being pushed right here together now what's interesting is if we look at these two traditions color correction has always been all about a gamma tone curve because it's in broadcast and as the name implies color correction we're just making little fixes before it goes live to air it's not about mastering or processing the image it's about like hey before that goes live you've got to trim a little bit of saturation or like oh man that contrast is just too heavy just pull a little bit out before it goes live but it's always done in a gamma tone curve and so that's where color correction practitioners are really comfortable living contrast that with color timers the way that if we go back to the days of uh photochemical film being the way that we captured and reproduced images which was most of motion imaging history until quite recently until digital intermediate that's a model photochemical film captures images by default in a logarithmic curve it doesn't capture them linearly so color timers but in contrast to color correction practitioners whatever you want to call that uh that that person are used to working on things in a log curve so that's kind of where we've been left today is we have these two warring disciplines one's used to gamma one's used to log there's all kinds of infighting and debate about like what's the best way to master an image because we've got these two very conflicted disciplines with different goals and agendas trying to ratify the best practices but what's interesting is neither of them really have any training or context for working in linear light unlike compositors who come from a more traditional kind of image science background you know most of compositing is really accelerated and began in earnest when uh digital compositing came into place i know there was an analog world of those things but as soon as compositors figured out oh we can do this in linear light of course we're going to do that they were much hipper to adopt that standard than than the the colorist tradition has been so that's a very long answer to what i think is a fascinating question of like what happened to linear and all of this why don't we ever talk about it or think about it or work within it and uh what what is the history that has sort of led us to that point so i'm not sure if you're here with us today clayton but uh maybe you can watch the broadcast later and get uh an earful about my personal hunch about why those things are the case let's see what else we got here okay difference between color boost and saturation good questions so let's just look at this in a really practical way and this is something i'll show you guys a trick that i use a lot when i'm wanting to evaluate questions like that which is all the time like we all have different approaches and different sort of how do you want to say it we all relate to our craft into our tools in a different way i have the compulsion to really understand what's going on when i touch a knob it's not enough for me to know that it's going to be pleasing or to get a pleasing result i generally want to know what is the math underlying that what is going on under there before i'm really comfortable using the tool and color boost versus saturation is a great question so let's look at our vector scope here for a second and let's just evaluate what happens with those two versions of manipulating saturation so saturation is a linear expansion and actually to make this even easier let's do this because right now we're kind of getting a free ride and a little bit of extra wrapping and uh unusual behavior or not unusual of desired but non-linear behavior out of this output transform so let's just work on this chart with nothing else applied for a moment and as you guys can see here saturation is evenly expanding all of my colors outward on the vector scope which is increasing the saturation by definition right because everyone knows if i set my sat to zero you can tell by looking at the image that i have zero sat but you could also tell me that by looking at your vector scope right nothing on here we're right in the middle bullseye so by definition we have no color all right so easy enough to understand that's what saturation does the color boost is there to try to account for the fact that generally or often i should say when we are going to manipulate the overall colorfulness or saturation of our image there are colors that are already quite saturated and others that are far less saturated so this the color boost knob is an attempt to sort of equalize those things and here's the simple logic when you turn your color boost in a positive direction to the right it is going to first sniff out the lowest saturations and it's going to crank those up at a higher rate than this then the colors that are already highly saturated so in an example like this one i don't know like let's see you know like this green apple is going to take the color boost sooner than the red ones to the right because they're more saturated or sooner than even better example like the blue bottle here and if we look at things on the vector scope you can see that sort of like characteristic it's almost like with the saturation uh version of the image let's go let's actually go back to our chart here with this version of the image when i add sat everything is just kind of evenly expanding outward right or it would contract inward if i drew it the other way if i go to my color boost it's going to feel more like you're pumping air into a balloon you guys see what i mean there we're kind of blowing it up and the lower saturations in the middle are getting ballooned outward while the outer saturations are not holding still but they're moving at a lower rate so what does all that mean practically when should we use it when we're grading it's very much a feel and a subjective thing it depends on what your scene calls for but generally if you take my premise that i'm always looking for the simplest possible tool so all things being equal saturation or color boost saturation because it's simpler it's a more linear simpler behavior so my criteria for saying well it's not the best tool even though it's simpler because i want a different behavior would be to be looking at an image where i'm saying gosh i'm getting the saturation i want out of this area of the image that was too low in saturation before however some of my higher saturations are now looking totally electric and crazy so that's kind of my criteria that i would apply apply for using saturation versus color boost and that's also sort of a a good way of introducing the sat verses or excuse me the the x versus sat curves here uh in this section of resolve these are other ways of articulating your saturation of your image in non-linear ways so we can say hey i want to saturate just the dark stuff and i actually want to simultaneously desaturate the light stuff or we can say hey the sat versus sat is actually a very good analog to the color boost it's going to do a very similar thing where it's going to say hey i actually want to desaturate the most saturated stuff and maybe while we're at it we will saturate the least saturated stuff and you can get as fancy as you want and articulate control points all the way along here and again that power is there to be taken advantage of and we absolutely should however it's worth noting that we should always be looking for the simplest possible solution right so if you can get it done with sat nob do it with the sat knob if you need the color boost use the color boost if you need to use curves use two points when you don't use three points if two will do you see what i mean so i hope that kind of answers the question there okay when should we use the 3d keyer the resolve training made it seem like a savior for the qualifier to be used where the qualifier failed is that so gosh that's a good question i i'm going to level with you guys i i'm going to tell you my least favorite tool in the entire resolve ecosystem it's the qualifier and the reason is i'm not saying i never use it because i do but the reason is if we go back to my concept that i'm wearing you guys out on of thinking photographically i try to think about the stuff that i'm doing like i'm back on set and like i'm an extension of rather than i'm a hairpin turn off to the left of the overall filmmaking process right so if i'm back on set and i'm looking at this image and we were to say hey it needs a little more uh contrast one way that i might look at introducing contrast and actually i'm going to cover this in the final part of the contrast series coming out next week but one way that i could look at that that would be a very photographic way of looking at it is if i create a window here and i decide i want to create some negative fill like so and then maybe drop this side down i give these to you guys all the time but this would be an example let's see if we can turn that ugly window off so it's easier to evaluate what we've got going on let me turn this off and that off and then we'll just be able to do an enable disable like so so that might not be all that we creatively want but you guys can see my idea there i'm introducing contrast in a very photographic way i could tell a gaffer hey let's get some negative phil in here on the phil side of her face so i'm always trying to look for those photographic real world analogs and when it comes to the qualifier what's the real world analog of a qualifier it's a very very like narrow and precise and very graphical way of grabbing at things that should be in my opinion your absolute last resort it has a lot of disadvantages i already made my analogy about the problem kids on the bus the qualifiers the problem kids for sure um and you know beyond that like to go back to the actual question of rgb versus hsl candidly i might not be the best person to answer it because i when i do keep qualifiers at all i still try to keep them so broad that i'm not interested in something narrower than the hsl qualifier can get me because i'm not even interested in anything i'm not even interested in getting like the narrowest version of what the hsl qualifier can get me and as best i understand it that's really where the rgb keyer comes in handy is when you want to get even narrower you want to get an even smaller subset of qualified pixels within your color cube that's where its utility seems to lie and that's just not something that really makes it onto my priority list generally speaking for grading uh will this stream be archived for later viewing well gosh i feel uh a a a little self-conscious because i'm just ripping with you guys here and uh you know i'm not not doing my normal scripted presentation but uh why not we'll leave it up and uh if everyone hates it then uh maybe i won't continue to do it but why not i'll i'll put it up for you guys to to rewatch at your at your uh pleasure uh let's see can we add any specific color in resolve with the specific hex code yeah you probably if i wanted to do that probably the way i would go about it is there's some kind of tool in os x for this it's the color picker maybe and in mac os i should say digital color meter is what this is called and i think you guys are just set up for uh you're only seeing my resolve if that uh i believe you're only seeing my result well point being to answer the question about uh hex values i would want to convert those hex values into an rgb triplet and remember that triplet is only useful if you know the color space is meant to occupy but setting that complication aside you could absolutely express any rgb triplet and resolve in fact i'll show you guys another kind of fun gizmo that i made recently for that kind of application let's look at my hsb generator here so this is just a triplet generator that allows me to use a couple different models oh interesting actually don't even have it doesn't even have an rgb mode but it should um but basically like you can imagine hue saturation these three fields instead of being labeled thus they could be red green and blue and you could just dial them in based on whatever input triplet you wanted to generate and of course you could do that i i this is probably even more complicated than we need to get i think you could even do that on the edit page with a generator yeah that was an over complicated answer to be honest solid color and then grab it click your color and you can probably even punch hex values directly in here yeah here you go okay that took me a second but that's a much better answer solid color click on the color generator and and uh key in your hex color right there okay let's see what else ah okay great question from uh nikhil i hope i'm pronouncing that correctly is da vinci why rgb color management better than asus cct this is such a great question especially because da vinci yrgb color management just got a whole facelift in resolve 17. so you probably won't be surprised to hear there is no single right answer to that question it depends on your needs and on your preferences however there are some key factors that we can consider when we look at both of them so let's talk about the pros and cons of aces first off so aces in the pros column asus has been around for a long time aces was the very first attempt at standardizing color management for motion images everybody's familiar with it it's supported across all kinds of different software platforms if you're working with different vendors or collaborators people speak it fluently generally if you say hey can i if i if i'm delivering assets to a vfx facility or receiving them back from that facility if i say hey can i give you aces linear or can you give me back ace's linear that's generally not going to be a problem because it's been around for a long time it's well standardized well documented so those are all kind of pros in the category for aces on the downside it has been around for a long time and it was the first person to the party if that makes sense they they have done a really good job of keeping it up to date but the fact is that it's built on a framework that's now well over 10 years old so there are aspects of it that i personally find to be less than optimal and uh the the other key factor to consider with aces is that as you guys who've watched my uh aces tutorial uh in the or the aces series rather asus is very turn-key even when we're doing it in the manual uh deployment that i was showing you guys in the asus series it's pretty much just three pieces of information what's your input color space what's your working color space you only have two options for that and what's your output color space that's it you plug in those three options and then you grade and that's really all you have available to you resolve color management by contrast let's just quickly look in our project settings and look at what resolve color management has to offer us if i skip over my presets and say no i want to drive i want to see what this baby can do and i go to custom look at all the options i have available here that's cool if i want options it's not so cool if i want a turnkey easy to use easy to understand model so those are some kind of kind of the key pros and cons there with resolve color management v2 versus aces and again i should point out with resolve color management v2 i actually am working in this uh ecosystem right now on a project that i'm doing and i'm using uh what do you call it davinci uh wide gamut intermediate really like it it's a really big color space there's lots to like about it however if i go back to those same vfx vendors that i was just talking about and i say hey can i hand you da vinci wide gamut intermediate files and can i get those back they're going to say no they're going to have no idea what i'm talking about so there's that aspect to be considered in terms of the interoperability and the standardization it's all very very new and it may not be well supported outside of resolve and that's the other thing is if you are doing any work in any other platform you really unless you're building custom tools or or going in and using the math that resolve provides or blackmagic provides for this color space you really don't have access to working mapping into or mapping out of this color space if you're not in resolve so a couple different uh ideas to consider there but there's no one that's better than the other and i guess the last thing i should point out as well is this also goes back to that that quandary that i mentioned early on today where even the simple question the simplest operation and color management of like okay what did my camera see what does my display need to reproduce and how can i mathematically transform between those that seems like a fundamental very simple question as i mentioned earlier not so simple because big container smaller container for just for the display so there's always going to be an aspect of subjectivity and a bit of a judgment call in terms of how we map down to our uh rec 709 display davinci yrgb uh or defend or resolve color management v2 as well as aces each have their own solutions to that problem that are technically sound but they also have a degree of judgment call in there because they have to so they produce different results so that's the other key factor for me is you should look at the results that each produces these are ultimately these tools are not really about aesthetics you should be defining your look yourself after setting up your color management but if you find as i often do that ace is for example like wow that's a very very strong contrast curve that doesn't leave me much room to add more contrast or add my own contrast when i'm grading in aces that sometimes makes me opt for working and resolve color management conversely if i want to have a nice silky smooth strong contrast curve out of the box and i don't want to have i want to do as little as possible beyond that sometimes that'll lead me to go more with aces so that shouldn't be your dominant deciding factor but you should consider that as well as like what is the aesthetic of that final output transform and is it more or less in alignment with my creative vision for the work i'm doing all right guys i'm going to do one more wow you guys have been awesome this is flown by and we've covered a lot of good stuff i'm going to try to hit one more question okay i lied i'm going to do two because we got two great questions here from uh from amandi and from zeke so let's start with almandie uh so what are your thoughts on placing grain at the start of your grade rather than at the end is there any difference that's a great question and for me the answer would come back to thinking about what i am modeling and that's generally you know grain has its own merits and we use it because it is uh has uh you know like aesthetic properties that we can find pleasing but i like to think about it in a practical way like all right what am i modeling here and specifically like the broadest question uh for me with grain would be okay what gauge of film grain am i modeling because eight millimeter is going to feel very different than super 35 and the next question immediately downstream of that is going to be am i modeling negative grain or am i modeling print grain and my answer to that question is probably going to be what determines where i place the grain in my node graph if it's print grain i probably want to place it where it would have been applied to the image in a photochemical workflow at the end of the stack if it's negative grain if i'm trying to cook that grain into the negative i would put it at the very beginning of the stack neither is right or wrong they're both going to yield different results because for example if you apply grain and then you add two stops of exposure that's going to give you a very different result than adding two stops of exposure and then applying grain so they're going to have different effects but i think a helpful way to navigate the differences and which you prefer is am i trying to model negative grain or positive print grain or am i trying to model a little bit of both in which case you theoretically could do a little at the head a little at the tail i've actually never done that before and now you've got my gears turning maybe i'll do that in a future project but that would be the way i would think about placement of grain in the node graph okay and then zeke's question another great one do controls work the same in aces as in rcm now or do they behave differently yet to use aces good question they're not identical because of that really of that factor that we just mentioned uh a moment ago of the way the output transform works because the tools are the tools no matter what right like these tools are not dependent on my color space they always behave exactly the same way exactly the same way and if you want to see what i mean go check out i can't remember if it's part one or part two of create better contrast but we do that you know very didactic like linear ramp demo of every one of these tools they always behave the same it's always the same math that's driving them what makes them feel and respond different is what's happening downstream of them what's happening in the output transform so as we just talked about aces has a different flavor of output transform than something like resolve color management so there's going to be variation in the response and the feel and the behavior of those tools there but generally speaking it's going to be quite minor compared to something like you know i i i mentioned in uh the last installment of the asus series this week that you know just fair warning if you're used to working in a display referred kind of by hand model when you're grading things are going to feel weird in aces because you're operating on the image in a completely different state so contrasted with a difference like that the difference of grading and aces versus grading and resolve color management to my hands is fairly minimal and then last one i'll i'll hit luke's here just on the way out so same question for noise reduction i assume that that that that's uh regarding uh placement in the node tree yep okay good i see it there there luke thank you uh yeah so another great question that i would try to think my way through before playing with and finding the right practical implementation of i generally want to think of noise reduction as a last ditch effort that i make last ditch isn't the right right word it's a finishing touch if that makes sense it's something that i'm doing as lightly as possible and for me generally as late as possible i want to hold off on doing something destructive like that for as long as possible until the very end and the other thing if you think about it conceptually is noise reduction is obviously a misnomer noise reduction is actually destruction of your image um there's no way around that it just sometimes yields an aesthetically preferable result so we should use it if that's uh if it gives us something that we want but it actually is destroying the the fidelity of your image a little bit it's blending and blurring things together in a way that doesn't totally kill the perception of sharpness so that for me is the final reason why i tend to want to place it at the end i want to do all of my work and give all of my tools access to the greatest amount of pixel data possible and then as a finishing touch only as needed when looking at the image in its final state just before it gets shipped to say okay let's add just a splash if i really really need it i know there are other colorists who feel very differently about that but that's my philosophy with noise reduction all right guys i'm going to have to call it here uh for today but this was really really fun i hope you guys enjoyed hanging out with me and uh let me know i guess in the in the comments on this video when i post it if you enjoyed this if you want to do more of them maybe we'll make it you know kind of a weekly or a bi-weekly thing i love talking about this stuff with you guys i love getting the chance to go a little bit more in depth on some of the better questions and comments that come through on the videos so let's get together again if everybody enjoyed it until then we'll uh see you guys for the next video see you for the next edition of grade school thank you

Channel: Cullen Kelly

Views: 14,999

Keywords:

Id: 6k6AZszhf64

Length: 64min 20sec (3860 seconds)

Published: Sat Apr 24 2021