Grade School: Using Parallel Nodes, Pro Colorist AMA

hey grade school what's up everybody it's Friday it's Friday morning or at afternoon or evening depending on where you are in the world it's one of my favorite hours of the week I'm psyched to be with you guys we don't need to be looking at my resolve quite yet you guys get to look at me in all of my sleep deprived Glory um I'm glad to be with you guys today uh I missed y'all last week I woke up like well actually no I didn't wake up I did this to myself on Friday morning I was working out felt a weird pinch in my back and for the rest of the day I couldn't really freely move without randomly getting like a spasm in my back that would cause me to audibly gasp so I decided to spare you guys that Vibe last week and uh to just wallow in my old Madness tell you something that that did not happen to me once in my 20s so life just keeps getting better and better um hope you guys are doing great it's a good day today we're gonna have some fun we got a multitude of things to cover from the last two weeks of uh YouTube content so I'm excited to dive in with you guys my buddy is here with me what's going on goodali oh hey I'm just uh winding down from a good trip to NAB still researching all the tech I learned about yeah you were the you you were the the Scout on that mission we still have to you gave me a a little bit of uh some of the takeaways but I'm excited to hear the rest of it because I didn't get to go and it sounds like it was a good show and there's some cool new stuff out there for sure I think um more about uh not a lot of as many new things as I had thought but there's a lot of important details that were good to pick up on so that was fun great to see the Blackmagic team as well and all their stuff they've got going on for sure depending on who you ask God or the devil is in the details perhaps they both are at any rate that's cool sounds like it was a good trip man glad someone got to go geez um all right well I don't have any remarks for you guys today I feel like I usually have something that uh is top of mind that I want to cover or set straight or show you guys I got nothing like that today so I'm gonna kick it to the room what kind of questions we got out there anybody wanna what do you all want to learn about today oh uh we've got a question or at least a comment perhaps from Lester saying I tried the composite mode you described in the video I wish there was a shorthand on what each of the items do some are obvious by scrolling over scrolling over others are not so I suppose the question is how do I learn what composite modes do what's the what's the math behind them yeah that's a great question to ask so this is one of those areas where we have to be sort of cautious when and how we use Wikipedia and certainly when and how we recommend Wikipedia but in this case if you literally Google in Wikipedia blend modes it will tell you the explicit math that underlies uh every one of the blend modes that you're going to find in resolve that really are all coming from uh the uh an institution that was developed from the Adobe ecosystem but that's where the easiest way to learn about it as you said like you can kind of pick off and start with uh the easiest ones like with you know multiply and divide those ones are pretty obvious and you know since we were talking about blend modes for a second I actually did uh a whole series on mixing light that it might be fun to sort of revisit the theme of here on the channel at some point where we did this series on mixing light that was called visual math by the way if you guys aren't uh on mixinglight.com or haven't had a chance to check it out uh it's a really really good resource that tends to have uh sometimes more structured material than we get the chance to do on YouTube so I definitely encourage checking it out in particular that visual math series if you want to get more of a practical sense of how I use different blend modes to do all kinds of zany interesting things but like as one specific example by the way I'm back here in good old safe and sound resolve 18 whatever it is 18 1 not 18 5 not the beta I I'm only Brave when it comes to exploring the Beta release so that we can talk about it on YouTube when it comes to my color grading business I'm quite cautious about bumping things or upgrading upgrading things before the dust has settled and I encourage that for you guys as well so I'm back here in 181 but we can still look at blend modes for a second here and we can look at a simple example of uh specifically what the multiply blend mode would do so for this exercise let me render my image visible so check this out I'm going to go into let me simplify a little bit I've got my template note tree here that you guys are accustomed to seeing now I've just got an empty node tree I'm within some color management but otherwise nothing else is going on I'm going to create a layer mixer because prior to 18.5 that's the only way to use a blend mode and I'm going to before I change my blend mode I'm going to go into my open effects here and I forget what this is called but it's like a color generator I want to say uh yeah color generator aptly named we're going to go in here and we are going to look at this color value and let's just set this we're on a scale of 0 to 255 right now for this value just bear with me I'm going to set this to 128 for all three channels so essentially 50 okay if anyone knows what I'm doing uh already then good on you because I barely know what I'm doing because this idea just now came to me so we've got everything set to a 128 which is 50 right 50 if the max is 255 255 divided by 2 would be roughly 127.5 but what's what's half a value between friends so if we look at our layer mixer right now we're in the normal blending mode which simply says look at bottom pixel and look at top pixel and top pixel wins every time right that's like the simplest blending mode that we can use that's the normal blending mode but if we wanted to use multiply I'm going to pose this question to you guys by the way I I think we we briefly touched on it uh two weeks ago in our last episode of grade school we just did CCA the colors career accelerator earlier this month month and boy was it fun one of the many many things that I find fun about the colorless career accelerator is that we get to interact I love interacting with you guys and fortunately we don't get to do a ton of it here in grade school because if I set the latency mode to low then the quality of the stream goes down and everybody complains and it's mad at me after the fact and I can't handle anyone being being mad at me I make enough mistakes as it is um so I won't have time necessarily to hear you answer the question but I do want to ask you the question and give you a second to think about the answer what is going to happen to this image if I set my composite mode to multiply right now give it to me in like what is the what am I effectively going to be doing or another way to think about it would be how could I do what you're about to see using conventional tools instead of resolve just think about that for a second and just because I love the interaction I'm going to open up this uh YouTube feed and maybe it'll actually be faster than I think probably not what it's going to do is apply a gain Factor oh there we go Elijah's on the right track we're going to apply a gain Factor so let's set this composite mode to multiply like so now are we gonna run into the good old-fashioned no okay good I thought we were going to run into the slightly odd behavior of the layer mixer sometimes when we're in a project level color managed setting so what are we doing here we have a constant that's what this generator really is up top it's a constant set to effectively 50 slightly slightly more than 50 percent and what the multiply blend mode does is say take bottom layer and multiply it by top layer like so so what does that mean well those of you guys who have hung out with me before have gone uh into geeky side missions with me talking about image math and the way that math is used to manipulate images and the way it relates to our primaries you know that multiplication and gain are two terms for the same thing right so effectively what we are doing when we multiply by a constant of roughly 50 percent is applying a gain factor of 0.5 so if I grab a still here and then wipe this all out let's see if I know what I'm talking about let's nudge this let's put it to a 0.51 to get as close to the factor that we were operating with before and let's see if we get a similar result her it's slightly different here so there is a multiplication going on there but why that's different I don't quite know because that should match right up that bothers me a little bit well regardless that's an adventure for another day it turns out we can find that gain Factor it could be a question of how that constant is being computed but it turns out a 0.79 gain Factor gets me a dead match to what we did with the multiply blend mode uh or at least very very close to it so maybe a question of like some other type of math that's going on under the hood but regardless if you want to get familiar with the math that underlies the blending modes that Wikipedia Google Wikipedia blend modes and that will give you that math and then of course you can just kind of naively experiment with things I'll be honest with you I don't have every single one of the like math functions uh memorized for all the different blend modes however especially with the release of that new feature that we were talking about in resolve 18.5 I have already started doing something at least for the day or two that I had 18.5 installed on a machine this week I've already started going in and being like I don't know try that one if it makes something interesting then I'll come up with an excuse for why it's okay for me to do it later you know that's my favorite thing about that blend modes thing is it really forces you into a realm of just subjective creativity and unless you are a exceptionally astute mathematic mathematics person you're probably not going to know what that blend mode is doing mathematically off the top of your head so all you you're left with is like well is it doing something interesting and now that becomes a sensible place to go like oh well what is that basis for that interesting thing anyway I've been going on about blend modes for a minute now but that would be how I would kind of bite off getting more familiar with the blend modes both on an aesthetic and a more mathematical basis question from Thatcher uh what operations are safe or unsafe in a parallel mixer yeah so that's a one that I definitely wanted to touch on today so this is I've been trying to think of a way to like simply explain this because it's a little bit slippery because even when I I think the terminology I used in my recent video was like oh spatial operations aren't so good but that's not really quite right because if I'm back in my template node tree here I can do a power window down below and you know whatever like maybe I want to do an outside window on my subject here and knock things down around her that's fine that's not causing any problem whatsoever and isn't that isn't a power window spatial by definition so here's the closest sort of concept the simplest concept that I've currently got in my mind in terms of being able to like uh explain it on a sort of programming basis the thing that we want to avoid doing in a parallel node is any operation which doesn't work on the basis of individual pixels so let me just give you an example of what that is and what that isn't so again for example we were just looking at gain this operation under the hood this operation is actually just being applied one pixel at a time so if we imagine the sort of like inner consciousness of resolve it's really only conscious when we are applying this gain operation of one pixel at a time it's doing what I ask it's literally taking like let's say you know like this you know like some pixel here and like the liner of uh her Mitten let's say for argument's sake that it is a one zero zero full red no green no bloat it's not that tidy but let's say that's the triplet all that we are doing is multiplying red times .92 green times 0.92 and blue times 0.92 and as a result in that case we're going to get 0.9200 because zero times anything is still zero um but my point there with the gain operation is That's Just Happening pixel by pixel so that pixel is going to get evaluated and have that math applied to it then it's going to walk one pixel over and do the same thing and then another pixel over from there and so on and so forth and it's going to do it 1920 times in the case of this canvas across and then it's going to repeat that sequence of a thousand nine hundred and twenty steps 1080 times so it's literally just going pixel by pixel by pixel that is a single Pixel operation any operation that works that way is cool and that's honestly most operations inside of resolve when I say spatial what I'm really talking about are operations where to perform the adjustment of any one pixel which is still happening the math needs to evaluate nearby pixels so the easiest example of this and the one that I made in the video on this subject would be blur the way that you blur an image is you look at the pixels around the pixel that you're currently evaluating so that's what I mean when I say spatial I mean the operation itself is having to know not only the value of the pixel it is currently manipulating but the value of pixels around it in order for the operation to work so a short list of those just kind of like running through here and resolve that would be blur that would be your texture pop tool interestingly that would actually be your shadows and your highlights inside of your primaries they're kind of an exception to what would generally be fair play here in the primaries because while these are a tonal adjustment they have a textural component to them they are changing the texture the relationship between the pixels as opposed to just the pixels themselves so shadows and highlights would be off limits for the same reason mid-tone detail would be off limits because that's making a textural manipulation and really texture is the the operational word that you can think of when you're deciding whether to do this uh we already talked about contrast pop noise reduction same thing that is a inter pixel adjustment it is looking at pixels to the left to the right above and below the current pixel any of those things anything with blurring sharpening noise reduction texture and as I said a kind of bonus that uh in a group that would generally be fair play would be the shadows and highlights just because those do happen to have a textural component now how do you know in the case of tools like that I don't actually know it probably is written down in the manual somewhere although I haven't checked in a little while but these are the two that I'm aware of Shadows and highlights that are not just like put it this way highlights are not just a narrower form of gain highlights are sculpting the top end of the image but they are also blurring and or sharpening the top end of the image depending on where that knob is placed so that would be my short list of things that you really we really don't want to be doing uh in the um parallel node structure because if you drive them hard enough you're going to end up getting that type of ringing that I called out in this past week's video it's possible that I'm missing something there or that my definition could be simpler or more complete but that's a pretty good road map for you if you keep those things offloaded Downstream of the parallel node stack over here you should be in pretty good shape switching gear is a question from Gonzalo uh what is the best uh what is the best way to practice for uh beginners in the colorist world should should they start with matching other movie clips and things like that or something else what's going to be the most helpful oh I'm really glad that we're talking about this I've been thinking about practice a lot lately here's what I would say one really good way to practice if you look at you know I was actually it's funny I was just talking with a colorist friend yesterday about how no colorist who I or they know came up in their career via a so-called traditional path so like the idea of saying there is a traditional path is sort of meaningless because I don't know anybody who had the sort of like well I worked in a post house and I was a runner and then I was a tape room guy and then I was a rotate room gal and then I was an assist and then I was a second and then I was or you know a junior and then a second and then a lead and then a senior and then a supervising I don't really know anybody who came up by that prescribed fashion that the kind of canned answer is like how do you make it in the color grading business however those people do exist they're just in an awfully small minority one of the advantages that people who do have that quote-unquote traditional career path one of the advantages that they enjoy is that they are typically given a very very finite set of parameters within which to accomplish a goal so I'll give you this example I can tell you uh pretty definitively that if you think about like any senior level supervising level artist at a big shop like company three or Harbor or picture shop any of these kind of places they are by preference going to be involved prior to production they're going to build the show Lut as we've talked about here on the channel before and they not only are going to get involved prior to production build the show let give that context for working on set they in many cases are actually going to be working with the dit and saying hey here is how I want you to grade this here are the tools that I would like you to use here are the tools that I don't want you to use and what I can tell you guys from my personal experience while I didn't necessarily have that full traditional path the time that I have spent working under senior colorist when at the time I was a junior or just a regular you know Garden variety colorist one of the things that I found the most beneficial there is when I was being a second on a show or a commercial or whatever for one of those senior colorists is it's their show right so they're not going to be bashful about like all right Colin you're here to help me out you're here to get me help me get through shots I trust your eye but you're going to do this by my book so they've got their look they've got the way they like to work they've got their template node trail out of the time and they have their tools that they like and their tools that they avoid and they will give me all that and that was some of the best practice I ever got was being told hey make a pass through this 30 minute episode of the show or whatever it is here are the tools that are in bounds for you everything else is out of bounds you may not use it so you can actually impose that same sort of uh structure on yourself if you think about it and one very sensible form of it that I think is amazing practice no matter how new or experienced you are as a colorist is to do what I'm always talking about here on the channel set up your color management set up an overall look whether that's my free 2383 or my Voyager or something else that you trust that is designed to work in a scene referred color managed space and just work through these three guys right here exposure ratio balance we just got finished talking about this in the colors career accelerator your goal if you're really serious about becoming an excellent colorist your goal should be that you're getting 80 by the time you've got color management look and these three nodes dialed in you're 80 done does that sound hard does that sound like a big bar to hit it kind of is that's why it's so important to practice getting to a really really good place with your images on a consistent basis and getting them to agree with each other using just these three notes so that would be one form of that broader sort of assignment that I got a few times in my career and that traditional colorists get lots of practice with of like hey go and grade these shots go and grade this timeline using all these tools these tools are off limits so that's I think the ideal way to practice as opposed to like this thing that we see a lot on YouTube this sort of like forensic comparison of like how are we going to make this random frame from my uh edit match this random frame from this movie that we like that to me is a there's a lot of problems there but it is certainly like you can call that whatever it is whatever you want but it is certainly not practice it is something else completely hope that helps Dimitri is wondering um what about the qualifier can we use that in a parallel node scenario yes yes absolutely because the qualifier that you know that I'm glad you're asking because it really is a slippery sort of Distinction but the qualifier is not looking at you know like remember the qualifier is making a qualification it doesn't need to know about the pixels around it in order to do its job let's imagine a one pixel or or um that's not a great analogy no the qualifier does not need to analyze pixels around it in order to do its job it just needs to analyze the pixel that we're that it's currently operating on based on its Hue sat and luminance and then we're just going to do whatever we do now the only exception there is if you marry the qualifier with those Outlaws that we talked about a moment ago so if you qualify and then you blur that qualified region problem if you qualify and then you do mid-tone detail or Shadows or highlights or any of those things that we talked about that's no good but if you qualify and you work your primaries or you do anything else that's inbounds then that's fair play and you shouldn't have any issue using it in a parallel node all right um let's see a question from Big Dog Films um would you please comment on the flavors of Rec 709 specifically what does it mean when it something is seen Rec 709a are those for acquisition or delivery yeah I'm glad we talked about this so uh we could literally spend the rest of grade school talking about the uh the the the intricacies and plot twists of the rec 709 standard and the way it has evolved but I'll try to put a tidy sort of thumbnail on it Rec 709 was it was originally a capture standard but even more significantly than that Rec 709 actually did not have a transfer function specified in it and that's what you're asking about right now any of the rec 709 standards that we are talking about I'm just going to give you a big long list right now of the ones off the top of my Dome Rec 709 gamma 2 4 Rec 709 gamma 2 2 srgb Rec 709a Rec 709 scene those are just the ones that I can think of off the top of my head there might even be others all of those color spaces they're primaries they're color gamut by the way was anybody in the color management Workshop that I ran with Blackmagic design earlier this week give a shout if you were we had a good time but as we talked about in that class as we've talked about here on the channel the primaries or the color gamut are one of the two ingredients that you need to know on an explicit numerical basis in order to fully Define a color space and of course that's what Rec 709 is a color space so all of those different Rec 709 color spaces have the exact same gamut exact same primaries the only difference between those two is the transfer function or the gamma curve or the tone curve whatever you want to call it and what I can say is if we work backwards from the knowns and the standards out to the slightly less knowns gamma 2 4 is the standard to which we generally grade and master our images sometimes called BT 1886 but BT 1886 in its simplest form is a pure gamma 2 4 curve okay and that literally gamma 2 4 if you want to get mathy about it that means you are taking your input value and raising it to the power of anybody remember exponents from high school I slept through high school I really wish I hadn't because I rely on math a lot these days but we are taking that input and raising it to the power of 1 divided by the gamma Factor so in this case it would be y equals x to the power of 1 divided by 2.4 okay so that would be a gamma 2 4 that's the standard that we all grade to gamma 2 2 exact same thing it's just 2.2 instead of 2.4 and as long as what you are sending and what your monitor is expecting a line that can work fine as well it's just a slightly different response those are the knowns it gets a little trickier when we look at srgb for example srgb is close to a gamma of 2 1 except it's not a simple gamma function and it has like a messy little toe business down in the bottom end so it's a slightly different encoding curve that I would generally encourage avoiding and staying away from and then if we look at some of the others Rec 709a is using what was sort of like informally agreed upon across the industry as the original curve of Rec 709 before we had a actual explicit standard for it that number happens to be around a is it a 1.9 we got some smart people in the room today someone correct me if I'm getting that number wrong it's a round gamma of a one point of a 1.9 if memory serves um so that would be the rec 709a now Rec 709 and for that reason that is not a standard that you should Master to there's generally not displays that have a gamma 1 8 response out there most displays are going to have a gamma 2 4 or a gamma 2 2 response or a response that's close to uh to one of those as opposed to being a 1.8 that would be quite rare then the last one that we could look at would be Rec 709 scene now I recently uh had cause to doubt myself on this one and I haven't gone back yet to check my understanding was that Rec 709 scene and Rec 709a are the same sort of roughly 1.8 ish gamma but I actually am not going to say that definitively right now because I feel like I want to go back and check I think the where I I heard someone suggest that Rec 709 scene is actually uh some other gamma function I don't remember it's a lot of a lot of tone curves so which of those should you use how much of all this stuff do you need to know as well or better than I do um I don't think you need to remember much of that stuff I think the important thing to say is we have to grade two standards we have to send what our monitor expects and then we can switch and grade to another standard if we want so one common example one simple uh way that we do things here at my company is my reference monitor is calibrated to gamma 2 4 I send Rec 709 gamma 24 to that reference Monitor and in cases where a deliverable is primarily going to be for the web there are a couple of things that I will look at doing but one of them that we do most commonly is we simply change that gamma 2 4 to a gamma 2 2 because we know that's going to be the most common response of displays that are going to view that content so that is far from a comprehensive and probably far from a tidy summary of all the different tone curves transfer functions gamma curves that we can encounter with Rec 709 but that's what I would really keep in mind is kind kind of like the takeaways like we have we we now do have a standard Rec 709 calls for BT 1880 or what it really BT 1886 is the proper term for rec 709 the tone curve that that color space calls for is a gamma 2 4 in the ideal simplest form so that's really what we should be grading to unless we know that we are grading to a gamma 2 2 type of standard and those others those might be useful for input transform so they might be useful for uh troubleshooting things but you really don't need to think about any of those other spaces nearly as much as is this two four or two two hope that's helpful and Tyler says when transforming log footage with csts luminance mapping washes out highlights I found drawing a custom curve and luminance mapping uh to match the log state is the best to bring them back but I'm wondering is there a better way ooh well you're touching on my favorite subject in the world which is curves I found myself repeating multiple times this week that God is a curve everything is curves and really what you're talking about can be most easily sort of understood if you look at a at a at a what we could call a parameterized curve meaning a curve that we're drawing with input parameters as opposed to freehanding with the custom curves so this will take a minute but this is valuable let's talk about let's talk about what uh is being called out here so what we're seeing when you identify this for a moment I'm going to bypass my color management okay and we can all see this right yep good now let's go in and do a color space transform and let's do the world's most useless color space transform I'm going to go from DaVinci wide gamut DaVinci intermediate to DaVinci wide gamut DaVinci intermediate this is a very exciting do nothing CST my waveform's not changing Scopes aren't changing images and changing right now all I want to do on top of this because it's all I want to evaluate right now is luminance mapping as we talked about and let's make an extreme example let's say we're going to go from a custom Max input of 10 000 down to a Max output of 100. this is what the question is identifying like oh my highlights are getting flattened or milked out right well yes but when we are using this properly this is so that we can fit into a domain that we need for display or for a that a color space transfer function so I'm going to give you an easy example that I just happened to know off top my head let's say we want to go to cineon has everyone ever anyone ever messed around with cineon before maybe you've seen that workflow where people suggest transforming into Rec 709 cineon to T up for a film look slot we're going to mess around with Cindy on here in just a second if I can remember what in the world I'm doing and find my grayscale ramp and what we're going to do now is we're going to keep all this junk the same we're just going to change our output gamma to cineon now check this out look what's happening my original signal is starting here it's just a nice clean pure linear ramp from zero to one but by the time we move into cineon which has a lower dynamic range those original bits some of them are crashing right through the ceiling right you can see we're hitting that ceiling hard so this is a case where even though if you're just grossly evaluating it visually you're like oh well it's just mushing my highlights if that's the best that your system can do as in the case of a tone curve like cineon then that's actually desirable because you don't typically want hard clipped highlights it would be better to do the minimum amount of luminance mapping that you needed to in order to fit this into the system and do not ask me why I know what that number is off the top of my head in cineon I've spent far too much time trying to emulate film Luts but you can see now we are rolling this in now if I only look at this on a visual basis and I look at luminance mapping off versus on in this case in this particular image it's not particular it's not super visible maybe if I zoom in here I would see it a little bit more off on yeah it's not even visible just because there don't happen to be pixels that are that bright in this image but it's something you certainly could observe in the right image and if all you're doing is looking at it visually then the conclusion you're that is being reached in the question is quite reasonable like well why would I want that why would I want to Mush my highlights you would want to Mush your highlights if you don't have a choice basically so that could be a tone curve like cineon which can accommodate all the dynamic range of your other one a more common one here array log C Airy log C or log C3 I should say it cannot accommodate 10 000 nits it can only accommodate up to about 5500 nits we can see that here as well so if we do not do any tone mapping here and we turn our ramp back on you're going to see same thing we're crashing through the ceiling so in this case we would want to go from 10 000 and we could go as 50 highest 5500 on the output and now we're going to fit inside of the available dynamic range of that destination okay so that's two examples in the abstract of just tone curves that need that but the bigger one the main thing that I use luminous mapping for that I think about luminance mapping for is for my display because essentially my display can only go up to if I'm in SDR it can only go up to 100 nits so even though I might look at this and if we sort of flush this like flush this out if I if I look at this I'm like oh I'm like flattening the heck out of my highlights it's like well it's either that or clip the heck out of your highlights because your display cannot show code values which come in higher than one and in this case that's what the this value of 100 correlates to is a code value of one all right so back to your original question now that we've got some context in place if we talk about like oh well I found that by stacking custom curves with my highlights uh or rather stacking custom curves with my luminance mapping I can get a better looking result I'm going to show you guys a tool that hasn't been shown outside of the colors career accelerator which we were just talking about but I'll share it with you guys here today let's actually look at that's not the one that I want we're going to look at this guy now this is a tool called Hilo trim it exists for the explicit purpose of trimming in highlights either on a creative or a technical basis really more on a creative basis and you can see that as I drop this down if I start to go to like a full negative 0.5 I'm really I'm most readily affecting my you know like sources in frame like this but it's also going to even begin to affect skin and stuff but here's what you can do and this is why I'm such a fan of Curves I'm I'm scaling you can see what my curve is doing right like here or there I'm literally just smoothly bending that curve down like so but check this out if I go to my roll strength param I can max out the luminance within this range so I'm still keeping things confined to a Max of this for whatever creative or technical reason I might want to but I can increase my shoulder strength and perceptually not just perceptually in reality get hotter highlights up to this point and if I go hard enough I'm getting dangerously close to really just having a hard shelf at the top of my curve but that's the phenomenon that you're witnessing by stacking a custom curve with luminance mapping is you are effectively increasing the shoulder strength of that luminance mapping that we have happening in resolve which by the way there's no parameter for in the uh color space transform inside of resolve there's no parameter for shoulder strength of luminance map that would actually be kind of cool if it was an option but we don't have it available so that's a very long way of saying good job if you're getting highlights that feel pingier and have more life while pairing with luminance mapping that's fitting you into the proper domain I think you've arrived at a very sensible and very advantageous scheme for getting the image that you want so I hope you guys enjoyed that uh grand tour of luminance mapping domain and all things curves God is a curve hashtag God is a curve Nate is wondering if you can explain uh ootf especially where it might concern hlg footage in an unmanaged CST workflow oh oh yes yes I can so let's talk about ootf for a second um ootf opto Optical transfer function that's catchy right ootf is a sort of companion to terms that are a little more common eotf and oetf Electro Optical transfer function and opto electrical transfer function those ones make a little more sense if we think about Electro Optical that's a transfer function that's meant to take you from code values to something visible right so that's like the gamma 2 4 we were talking about in Rec 709 earlier that's an electro Optical transfer function that's how we're getting from code values to something visible opto electrical transfer function that would be something like log C3 that's how we are getting from something Optical visible out in the real world into an encoded value and the transfer function describes the way that we map those luminance values as they climb in strength okay so those are the more kind of easy ones to understand so if we now know what an eotf is and an oetf is what the heck is an ootf we're going from opto to opto opto Optical transfer function what does that mean the more common term for this that I was familiar the only term for this that I knew before ootf was introduced in resolve 15 or 16 somewhere around there with system gamma it turns out that for a variety of reasons like I just want you guys to imagine like everything that we do in motion Imaging like let's let's use my my beautiful mug as an example here I'm standing in a physical space in an actual room right now and I'm looking at a camera lens and the Hope fingers crossed is that you guys on the other end are seeing a linear reproduction of what you would see if you were standing where my camera lens is standing right that's kind of the whole Baseline of motion Imaging without that we don't have much so in that scenario the most naive uh sort of like prescription that we could give would be well in that scenario we would want an opto optical transfer function an end-to-end system gamma if you like of one right don't change it show me what it looked like to be standing there as it turns out because of the limitations of displays because they don't have as much dynamic range because we are looking at those displays in different levels of environment with different levels of ambient light that are causing Flair on the screen it's actually advantageous to have a system gamma which is greater than one so that you have a little bit of additional contrast in the final image there's no hard prescription for this but generally speaking you could we could go back to our rec 709 conversation and say that a very common system gamma would be I'm doing that the wrong way actually let's look at a 2.4 divided by 1.96 as I saw a couple of you guys uh fine-tuned my uh my my value on a moment ago 2.4 divided by 1.96 that's around a 1.2 so what that means is that all things being equal if this is the original scene what I'm looking at right now let's let's actually let's let's build this thing let's do something here so I'm going to do this with a CST for a moment instead of I've got my color management bypassed and I'm going to going to just do some simple shot level color management for a second here just want to make sure all my stuff is off Okay cool so we're going to do an output of DaVinci wide gamut to 709 using a color space transform and we're also going to do and actually let's just make those really simple we're just going to do it in one step we're going to go directly from camera which is log C3 straight out to rec 709 gamma 2 4. okay here is my CST and as you can see I've got this forward ootf turned on let's grab a still of that let's turn that off and this is not going to be exact because of where that ootf is being applied but if I now go in and I apply a numerical gamma Factor using a tool that lets me do that and I put in a gamma of around a in this case it would be a whatever that original value was of 1 divided by 1.2 around a 0.81 like so I'm gonna get in the neighborhood of the system gamma of that ootf so the ootf is really just a way of prescribing a traditional system gamma and I think it's I think it actually is a 1.2 it just happens to be applied at a slightly different point in the image than I'm applying it right here but that would be the idea there is the ootf is the end-to-end gamma of the image now to round out the answer you're asking about hlg footage that you were then reading back in well this is where working with hlg as an acquisition format is kind of tricky because when you think about oh and something just clicked in my brain as well that I'll come back to when you think about hlg hlg is a display standard which means it probably has had some type of system gamma imparted into it at least if it's like if I say hey here's the hlg master of you know like Cullen's new Hulu series that I've just gone out and made it's going to have a system gamma apply to it because it's meant it's meant to have slightly more Gamma or contrast than the original image did but if I'm giving you footage that was shot in hlg it I don't actually know maybe you guys know in the room it's not entirely clear to me whether I should expect that system gamma has been applied to that image or more likely that hlg is simply being used as a reasonably efficient and dynamic way of encoding a source image in a like camera negative high quality State I would expect more of the latter in which case you would not want to use ootf inverse or forward when you are like input color space transforming when you're color managing that image but I actually don't know that for sure and you might have to go as naive as trying both and seeing which one generally looks better on your material for a particular camera but that's kind of the rub with uh how I would treat that hlg application in particular and just in general here's a good rule of thumb if you don't remember one of think about all that junk that I just said about ootfs these two options that we have down here these come into play when we are in two situations we are going from a scene or camera space into a display space that's what we're doing here we're going from area wide gamma 3 area log C3 into a display space or we are going from a display space into a camera space such as area log C3 DaVinci by gamut intermediate whatever it may be in that case we may we may not but we that's where one case where we might want to use an inverse ootf and that first situation that I uh summarized of going from scene to display we're almost always going to want to use a forward OTF because system gamma is generally our friend it's a pretty well agreed upon convention after you know a century or so of it happening we have system gamma actually all the way back in film days and in that case you're talking about a very dim image in a room with very controlled lighting and the system gamma it turns out of a film print is going to come out to I believe top of my dome right around that same value of about a 1.2 so there you have it more than you ever want to know about ootfs sticking with the topic of color space transforms your server is asking should I use white Point adaptation for going from DaVinci what gamut to P3 well that's a good question to ask because P3 has more than one sort of conventional white point that you might that a display or a standard might want depending on what you're doing with it Rec 709 you know like there's there's really not a Rec 709 like D60 standard for example or a Rec 709 DCI standard so you generally we just know when we do Rec 709 we're going to want d65. for P3 it kind of depends on what you're doing and you can even see it if you look at the color space we've got p3d60 p3d65 and P3 DCI so you want to make sure I would say if you have if you are confident you have selected the correct P3 of these three options these are white points by the way D60 is slightly warmer than d65 d65 is what we set our reference monitors to it's what most color spaces are pointed to like that we're going to generally be working with DCI is a projection standard which is like slightly greener in character so as long as you are confident that you have selected the right P3 the one with a white point that reflects your needs then I generally would turn a P3 at it or excuse me white Point adaptation on otherwise effectively what you're going to be doing like let's paint this with some specifics just to make it a little bit easier let's say we're going to do a P3 DCI gamma 2 6 DCP for example okay something for a theatrical projection we'll say this is going to go to P3 if I do not turn my white Point adaptation on here what I'm going to be doing is simulating the white point of my source on my destination so even though on this hypothetical projector that I'm sending that I'm mastering this image for a triplet of 1 1 1 equals the white point of DCI I'm now going to be sending my white through the system not at a 111 but at the code value which will cause that projector to reproduce the same white Point as my source if that sounds slippery it kind of is it probably took me two dozen times of having someone explain that to me before it kind of clicked so that's a very long way of giving an answer that I could very briefly give of saying yes if you're going into a P3 space and you know you have selected the right flavor of P3 then use white Point adaptations so that your Source White is being mapped onto the white point that is native to that device that you are mastering for moving over to contrast Praveen is wondering how do you play with pivot to find the right amount of contrast is keeping it at the right amount for the working color space not enough oh yes I'm really glad this one has been coming up a lot lately and I sort of feel I I really want to get on the record loud and clear as often as I need to for you guys like I can be very prescriptive in the way that I talk about what we do and I really do believe that you guys have probably heard me share that like David Fincher quote that like there are not a million ways to do things there are two and one of them is wrong like I really do feel that way so much in my practice and I do believe in Optimal Solutions to things and I try to impart that to you guys but I also uh have tried to be very consistent but I and want to continue to emphasize to y'all that by the time we get let's just kind of go back to a stock grading setup here we've got color Management in place I'm going to go in and pick out a tasty Voyage a lot so I got a little bit of a Vibe going on my current favorite is this Canberra thing I'm going to scale back the key output gain a little bit by the time I am here there is nothing and I mean nothing sacred about the pivot of my contrast ratio it is all about what feels good in my hands everything that I've done up until this point has been done in the name of giving me a level position to grade from so that when I start to work contrast pivot or exposure I'm not inadvertently having to compensate for bumps that I made in the pipeline when I set up the color management or the overall look but now that I'm here the only thing that matters is how I feel and how my client feels so I don't even I don't know what pivot I land on for most shots because I just grab these two knobs and twist them until I like what I'm seeing so I I not only would I say it's not enough to Anchor at middle gray I'd say that's often not going to be appropriate like contrast and exposure are interrelated and even though I talk to you guys about like well if you set your exposure then you shouldn't need to work your pivot as hard you know set that where you want it to go and then just feel free to like work your contrast and your pivot I never have a hand on just one of these knobs I always have a hand on both because I'm just honestly looking at the image it's a variation like contrast pivot you guys will know this if you've heard me talk about these the math Behind these things in the past contrast pivot that's really just an alternate form of lift and Gain it's like literally the exact same math just in a different form and same thing like I would never go in and make a lift gain kind of contrast feel based adjustment and take one hand off the wheel they interact with each other they're actually supposed to it's a good thing same thing with contrast pivot but that's a long way of saying like by the time you are grading contrast ratio at the individual node level there is no forbidden pivot point and here's something else that I will do like I do this all the time if I want to make a gain adjustment usually what I will do is I will like before I even start doing stuff like if I this image is actually a decent example like let's just reset this node I would look at this image and be like all right by the time we've exposed up like let's say we wanted to go a little more open with the exposure I could definitely see myself looking at this image being like you know what like the overall weight feels pretty good it just needs a little bit of a broad Trim in the highlights before I even turn my contrast I'm going to set my Pivot to zero and then start dropping down what am I doing there answer to yourself if I'm doing a contrast adjustment where I'm pivoting around zero what am I doing assuming that in this case my s-curve contrast is set to off I'm just doing gain right but what's fun is if I feel like I'm doing game but I want to keep a little bit more I I want to like finesse that a little bit this is where there's an advantage over you over using gain when we use contrast pivot because now I can make that pivot just slightly higher than zero without needing to change the adjustment over here and now my adjustment is taking a little bit of contrast a little bit of weight out of the bottom end as well even though it's taking most of the weight out of the high end because my Pivot is set at a rather low value so I hope that's some helpful just sort of like I want to give you like the blessing the creative freedom to go and twist and turn contrast pivot wherever you want them get a good color managed pipeline get a good look in place set exposure where you want to see it and then wherever your hands Lead You In the contrast pivot or in your ratio node in general you should feel uh totally free to go there all right moving over to Output BG is wondering should we set the output to srgb if we want to deliver to Instagram YouTube or the internet in general no um I I want to link you guys to a video uh that is really good uh that one of my do you guys know about my I'm sure some of y'all in the room know about my color science crushes I've got a few color science crushes uh out there one of them is a really sharp dude uh at film like called uh Danielle sergosano Daniella has a video that I'm going to try to dig up for you guys in post where he breaks down the whole gamma tag and uh you know like srgb standard I would say just avoid srgb uh entirely because even if a display's actual this is the crazy thing about srgb and I I'm sorry that there's not a simple way to explain this the beauty one thing that's really nice about a gamma function let's just talk about Gamma let's talk about the simple version of this first if I tell you I'm grading direct 709 gamma 2 4 what that means is that I am applying a gamma 2 4 function to my image on its way out the door right and what that also means if I'm using it correctly is that my display is going to do the perfect inverse of that so that the net result is what linear nothing changed right just pure linear in and out setting aside the ootf thing that we were talking about earlier it's just gamma 2 4 on the output gamma of 1 over 2.4 uh in terms of the display response so the net is linear same with gamma 2 too it's it is to use some fancy terminology it is a symmetrical encode and decode srgb depends on who you ask but it is not guaranteed to have a symmetrical encode and decode some standards some displays some experts will say until they're read in the face that the srgb standard calls for an encode of an ugly little function that you can look up if you really want to see the math it's an ugly compound function that can't be described by a simple gamma curve and then a decode of gamma 2.2 then others will say no no no no you should be decoding the exact inverse of that forward function but there is not agreement there and I've heard enough smart people debate it on either side that I don't want to get involved I don't want to deal with that at all so that in and of itself for me is the reason why I'm like just avoid the srgb thing just do gamma 2 2 because even if you do srgb flip a coin as to whether it's going to be decoded as srgb or decoded as gamma 2 2 so at least half that time you're going to be better off or no worse off than if you had selected gamma 2 2. I I know that's complicated stuff but again I'm trying to give you guys like thumbnails at the end of each of these skip srgb if you're trying to get a more robust transfer function or gamma curve for web use gamma 2. all right um let's see a couple questions one from one from Sean and Maximilian uh I've seen conflicting information on using iPhone footage in a DaVinci web gamut intermediate project uh what's your process for color managing that and Maximilian chiming in saying how do you manage projects with footage that can't be set in the color management like f-log two for example oh yeah two really good questions so for the iPhone stuff I would refer us to the grading Rec 709 footage uh video that we did not that long ago here on the channel basically put it this way with iPhone or with any other sort of seven or nine baked material by the way what we were talking about earlier like that big digest that we did on Rec 709 gamma 2 4 gamma 2 2 Rec 7019 right 709a srgb all that like that Insight that we called out of like oh those actually all have the exact same gamut the exact same primaries one of the times when it's useful for us to know that is in situations where it's like how am I going to deal with this iPhone footage because we at least know the gamut of that iPhone footage the gamut of that you know mystery meat material that you're being given for your documentary or whatever it is that you're working with the gamut of it you know it's gonna almost certainly be Rec 709 or srgb those gamuts are the Same by the way you that part you you know so it's really just a matter of fitting the tonality and getting you the best starting point for your grade so as we talked about in that video the options that I would look at would be as follows you have let's just pretend this is my input CST for a second year so I would look at my input gamma that could either be Rec 709a gamma 2 4 or gamma 2 2 okay those would be kind of the input Gammas that I would look at and then you could look at whether or not you apply your inverse ootf those would be the variables that I would play with and I would play with them very naively and just try to get the best looking starting point for your image all right so that's number one there and then the second question jog my memory Goodale uh that's about dealing with um formats that aren't in resolved right right yeah so yeah we're asking like what do you do when you have a format that isn't supported and resolved I tell you what I've started doing that I'm really having fun with I go uh and talk to my community and say hey can we find the standards for this and can we build something that allows us to accommodate that input format and I've been teasing this to you guys here on the channel for a while but I promise we've got a large set or a significant set of input transforms for unsupported cameras that we're going to be releasing here on the channel for free very soon because it sucks when that comes up and what I've done in the past is either build myself or ask someone smart to build an input transform based on the published math of that color space if it can be found oftentimes it can be and it's just a question of there being a bit of a lag between it being out there and resolve implementing it as a color space but that's my first uh Choice there the other thing that I will will do this is a I'll I'll sort of cautiously give this prescription this advice if it's log format X and it's either an unsupported log format or if you're like this is clearly log but I don't know what flavor of log it is something else that you can do that I won't tell on you for if you make a good looking image come out the other end is you can bypass input transformation completely because remember DaVinci wide gamut DaVinci intermediate that's a log grading space and if you don't know your input transform or you can't find one that feels like it's getting you to a better position just skip the input transform and grade through the issue that's I I feel conflicted giving that advice because that's not very pure that's not very scientific but sometimes that's the best you can do and it can definitely work so that's another option to you as well like if you got to leave grade school today and go grade a piece of content along those lines if you can't find an input transform that is some flavor of log and some set of primaries that you can kind of reasonably guess to then just skip it and know that you're probably going to have to decrease your saturation a little bit and you're going to have to sculpt the high and the low end of the image to make sure that you're not clipping values out just another another option as well all right we got time for one more okay um let's see I gotta make it a good one um all right question from Alexander uh how do you set up your look first at the Timeline level and go back and do foundational work on a clip later I think the question is more about um what if the look is applied to something that isn't correct in the first place yeah this is another one that I actually really like how it's been coming up in slightly different forms a lot lately and I like it and I also like the the tone that I'm inferring from your question like how are you going to go and do that how are you going to like have a shot that's out of balance and then establish a look based on it and then grade it and then expect that the look is going to travel on it that's a good question here's what I would say this is a difficult thing but a really valuable thing to keep in mind the the another form that this question has often come to me in in like the last couple months is like well dude you're kind of giving yourself softball pitches because everything that you grade on your channel and everything that you grade in grade school has a certain level of quality to it and it's pretty consistent and meanwhile I'm out here in the weeds and I'm grading projects with like Bunches of different cameras mismatches and white balance mismatches in lighting like big problems and the footage is all over the place and it's not feasible for me to set a look before I do shot level balancing on those shots so how are you going to encourage me to do that like that's been another common question I've been asked here's what I would remind you of it's difficult to see sometimes but I don't care let's take 10 shots and say these are the 10 most unique shots these are 10 shots that are like 10 they're they're a 10 pointed shape in space they are like the maximum distance from each other not a one of them looks anything like any of the other nine okay like the extremist most far-flung example you can think of of 10 shots that don't even belong in the same edit okay that's like literally the worst case possible scenario even in that scenario there are common characteristics you just have to find them and even in that scenario there is a shot which represents the median of ranges of contrast ratio ranges of exposure ranges of color balance there is a shot which rests in the middle of all those others and all the others are kind of radiating out this one's higher contrast ratio this one's more pink this one's got a lower exposure but there's one that represents the middle and it's not always easy to find but it's worth looking for that's the one that you set your look on and then you match or balance your other shots like alongside your look Dev to line up with that shot so it's not crazy it is not wrong to say like I don't have a great example in this particular timeline but it's not crazy to set a look in on that Bullseye image that best represents the average or the middle of where the footage is coming into you at and set a look there and then you bounce over to shot number two and you're like yeah but shot number two is cooler well so flip over to your timeline or to your clip level grade rather and warm it up a little bit and now go back and evaluate okay now between my Baseline shot and this new shot does the look work and do the two kind of agree with each other so there is an interaction there's a push and a pull but the biggest thing I would remind you of is that even when things feel like they are completely disjointed and unrelated and that there's no there's no setting of a global look before you go in and grade each individual shot to align with all of its neighbors there is a shot that represents the average the middle the center of uh like range of exposure contrast ratio and color balance you just have to find that shot and start your look Dev process there and then from there you're going to match that or evaluate that look on other shots and you are probably going to need to flip over the clip level and finesse your exposure contrast ratio and balance so that that shot is a bit more in line with your Center Bullseye shot that you started with I hope that's helpful it's I I understand the question and I'm really glad you're asking because it's a a real one and you're talking about like one of the key parts of this craft to me anyone can do look Dev when all the shots agree with each other and they feel really finesse and tight it's a lot harder when you've got a more far-flung data set like we're talking about um that's a great one to end it on wow that was a fast hour guys y'all are awesome always enjoy hanging out with you guys on Friday morning um I don't think I have any announcements oh by the way did everybody check out the color code the podcast came out last week my interview with Andrea cleback from Arbor Picture company dang that was cool she is I told her at the beginning of the episode and I 100 met it she meant it uh she is one of my absolute favorite color was working she has such a diverse thumbprint as an artist and we had an absolute blast chatting so go check out that episode if you haven't already definitely worth every minute of the whatever like 80 minutes that we spent chatting together we got more of those coming very soon so if you didn't get a chance to go check out uh the channel it's a video podcast too so there's the color code here on YouTube it's on Spotify Apple podcasts all those different places uh go check out that podcast that episode with Andrea and the forthcoming episodes uh and you guys can let me know uh I don't know let me know in the next Community post maybe I'll put up a poll for you guys what colorist what motion Imaging artists would you want to see me sit across from and talk with for an hour I've got a list that stretches from here to the next block but I'm always looking to add more to it so let me know and maybe we can make that happen I'll leave it there for today thank you guys very much for joining me for uh grade school today awesome session as usual have an awesome Friday and an awesome weekend and I will see you guys here on the channel very soon take care

Channel: Cullen Kelly

Views: 4,081

Keywords:

Id: SYDiFTnZiCA

Length: 65min 4sec (3904 seconds)

Published: Fri Apr 28 2023