Jump to content

35mm photo film emulation - LUT design


hyalinejim
 Share

Recommended Posts

I haven't been active on the forum very much recently, but I've still been reading and following discussions with interest. The main reason I haven't been around much is because I've been busy trying to emulate the colours of 35mm film for digital. I've focused on contemporary photo films (as opposed to motion picture film), trying to get as close to filmic colour as possible in Lightroom/Camera RAW. I've got quite good at it now: I can get a pretty accurate match. And I can use the same method to create LUTs for video that are also pretty accurate. However, post processing photos is a bit different to colour correcting video and I'd like to ask the forum's help in trying to figure out a few things in the LUT design.

(To date, I've only tested my technique on GH5 VLog, as that's what I shoot. But the LUT should work on VLog from any Panasonic, although colours will be slightly different as each model has a slightly different colour response.)

There are a few different components to my method:

1.  Match overall contrast
2. Match hue and saturation of colours
3. Match colour shifts in the shadows and highlights

Here is an indicator of its accuracy. On the top is VLog with my LUT and the bottom is the film scan (Portra 400 exposed at +1 stop)

02.thumb.jpg.393e5b5db48552cb6dff4b849045f95b.jpg

03.thumb.jpg.4c1762b6bf45b4c5277680f1162b3799.jpg

Here's a link to a folder with the LUT. It's for Panasonic V-Log and emulates the contrast and colour of Portra 400 overexposed by one stop and scanned on a Noritsu minilab scanner with slightly reduced contrast. However, I've removed the colour cast in the shadows and highlights for the sake of keeping the discussion simple, for now.

https://drive.google.com/drive/folders/18RiY7dZ6AO87yMKArUvRDNKfouUVCu8H?usp=sharing

My first question for you guys is what do you think about the level of contrast? I've actually reduced it a little bit relative the scanner's default contrast level. But is this the right level of contrast for you?

Even though I've lowered the contrast level a little bit, it might still seem quite contrasty. That's because it's not designed as a "wide dynamic range" LUT that preserves much of the log signal, that you would then do further work upon. It's designed to give a "Rec709" level of contrast, so about 7.5 stops (which has traditionally been accepted as the average dynamic range of a scene). That means if you have a high contrast scene you might need to lower the contrast of the log signal before the LUT. Furthermore, it's designed so that middle grey as shot equals 126 RGB in Rec709 (almost 50 IRE) when the LUT is applied, which is what it should be. So any under or overexposure in the log footage will be quite apparent and require an adjustment before the lut. Basically, exposure and contrast and WB corrections should take place before, not after the lut. And, of course, it should be applied in a Rec709 colour space. If you can find a reliable way of making those adjustments directly on the log footage (you can in Resolve and After Effects, you can't in Premiere) then grading becomes very quick and easy.

Here are a few VLog stills with the LUT applied to show how it handles skintones, saturation and contrast.

001.thumb.jpg.3f2de1f47a6915944cdd357af4ce25a5.jpg

002.thumb.jpg.08f316b0ba3043986df73caa58ad33f9.jpg

003.thumb.jpg.38cf44accc4c221842da3b2fc21ca6c3.jpg

004.thumb.jpg.bcfa4a1ed29eefb470722c18911d6770.jpg

005.thumb.jpg.f12538627394f0821510d9706c8d5a4f.jpg

006.thumb.jpg.939bf24866d34915925b2a356853e20e.jpg

007.thumb.jpg.af3d26c1d7f3a08c8fa20328ba90745e.jpg

Link to comment
Share on other sites

EOSHD Pro Color 5 for Sony cameras EOSHD Z LOG for Nikon CamerasEOSHD C-LOG and Film Profiles for All Canon DSLRs
10 minutes ago, webrunner5 said:

Yep it looks like a GH5. Hmm.

Let's take one aspect of the LUT which is possibly its most important: hue, saturation and lightness of colour and let's leave the contrast constant and compare the film emulation (bottom) to Panasonic's own V-Log to Rec709 colour transform (top). You might agree with me that one of these looks better than the other in terms of luminance and hue of skintones:

1511665211_P1022130(0-31-21-44).thumb.jpg.ff50d893ee490643a608bc6049c91896.jpg

502869803_P1022130(0-31-21-44)_1.thumb.jpg.e252bd4424bef280f91c21d411ec21f0.jpg

Portra 400 is famous for its attractive skintone reproduction, offering slightly desaturated, even porcelain-like skin which skews more towards tan than pink in terms of hue.

Link to comment
Share on other sites

2 minutes ago, webrunner5 said:

Yeah but you could probably under expose the top shot and get close to the bottom one. More than one way to skin a cat as they say.

There's more going on here than cat skinning. Try the LUT and you'll see.

Link to comment
Share on other sites

No your just making your life more boring. I grew up with no LUTs. Oh how did we ever make anything look good or even great.  Magic I guess, or maybe a little pre thought before shooting. Like lighting, makeup on and on, but sure things have changed but I think for the worse, a bit of ahh hell just fix it in post mindset. Comes off as a bit of laziness me thinks. Lack of imagination to boot.

 Not really picking on you. I can see you have put some time into doing this. I just am beginning to see all the cameras looking alike and I think NLE's have a lot to do with it to be honest. Not liking the trend.  Blackmagic cameras seem to be the only ones I like anymore. They have bucked the trend.

Link to comment
Share on other sites

Oh honesty doesn't count. And you are going to sway me that a GH5 has some super nice Color Science to start with. Interesting. I have been doing this a Long fucking time. I sort of have a clue. But yeah done commenting. Lead on. Not too impressed.

Link to comment
Share on other sites

3 hours ago, hyalinejim said:

I haven't been active on the forum very much recently, but I've still been reading and following discussions with interest. The main reason I haven't been around much is because I've been busy trying to emulate the colours of 35mm film for digital. I've focused on contemporary photo films (as opposed to motion picture film), trying to get as close to filmic colour as possible in Lightroom/Camera RAW. I've got quite good at it now: I can get a pretty accurate match. And I can use the same method to create LUTs for video that are also pretty accurate. However, post processing photos is a bit different to colour correcting video and I'd like to ask the forum's help in trying to figure out a few things in the LUT design.

(To date, I've only tested my technique on GH5 VLog, as that's what I shoot. But the LUT should work on VLog from any Panasonic, although colours will be slightly different as each model has a slightly different colour response.)

There are a few different components to my method:

1.  Match overall contrast
2. Match hue and saturation of colours
3. Match colour shifts in the shadows and highlights

Here is an indicator of its accuracy. On the top is VLog with my LUT and the bottom is the film scan (Portra 400 exposed at +1 stop)

02.thumb.jpg.393e5b5db48552cb6dff4b849045f95b.jpg

03.thumb.jpg.4c1762b6bf45b4c5277680f1162b3799.jpg

Here's a link to a folder with the LUT. It's for Panasonic V-Log and emulates the contrast and colour of Portra 400 overexposed by one stop and scanned on a Noritsu minilab scanner with slightly reduced contrast. However, I've removed the colour cast in the shadows and highlights for the sake of keeping the discussion simple, for now.

https://drive.google.com/drive/folders/18RiY7dZ6AO87yMKArUvRDNKfouUVCu8H?usp=sharing

My first question for you guys is what do you think about the level of contrast? I've actually reduced it a little bit relative the scanner's default contrast level. But is this the right level of contrast for you?

Even though I've lowered the contrast level a little bit, it might still seem quite contrasty. That's because it's not designed as a "wide dynamic range" LUT that preserves much of the log signal, that you would then do further work upon. It's designed to give a "Rec709" level of contrast, so about 7.5 stops (which has traditionally been accepted as the average dynamic range of a scene). That means if you have a high contrast scene you might need to lower the contrast of the log signal before the LUT. Furthermore, it's designed so that middle grey as shot equals 126 RGB in Rec709 (almost 50 IRE) when the LUT is applied, which is what it should be. So any under or overexposure in the log footage will be quite apparent and require an adjustment before the lut. Basically, exposure and contrast and WB corrections should take place before, not after the lut. And, of course, it should be applied in a Rec709 colour space. If you can find a reliable way of making those adjustments directly on the log footage (you can in Resolve and After Effects, you can't in Premiere) then grading becomes very quick and easy.

Here are a few VLog stills with the LUT applied to show how it handles skintones, saturation and contrast.

001.thumb.jpg.3f2de1f47a6915944cdd357af4ce25a5.jpg

002.thumb.jpg.08f316b0ba3043986df73caa58ad33f9.jpg

003.thumb.jpg.38cf44accc4c221842da3b2fc21ca6c3.jpg

004.thumb.jpg.bcfa4a1ed29eefb470722c18911d6770.jpg

005.thumb.jpg.f12538627394f0821510d9706c8d5a4f.jpg

006.thumb.jpg.939bf24866d34915925b2a356853e20e.jpg

007.thumb.jpg.af3d26c1d7f3a08c8fa20328ba90745e.jpg

I don't have V-Log so can't apply it my my own footage, but I have a few thoughts.

To me, I think the colour charts can only get you so far with skintones as they don't have enough hues to work with.  Assuming you're shooting your own film and can point it at whatever you want to, I'd recommend making your own colour checker that has a much more detailed set of hues around the skin-tone hues.

You could either do this the arts and crafts way, and mix some paints.  I'd recommend getting a hue that is quite red and one that's quite yellow and then blending them together in various proportions to create a number of steps that smoothly go between the two hues.  if you then mix each of those with white in greater proportions you should get a version of each hue in gradually diminishing saturations.  I'd let them dry then compare them to your own skintones and see if they include your own skintones.  You might have to try this a few times to get the source colours and strengths right, but you should be left with a grid that covers the pie in the vectorscope around the skin tone indicator quite well.  

The other way to do it is to do it digitally and then have it printed.  Then try and compare it to real skintones and iterate if required.

Those tests won't be calibrated the way a colour checker is, but it'll give you the ability to really dial in the response of that critical region, which is useful as long as you can shoot test shots from both cameras under identical conditions.

The other dimension is to shoot the charts in a number of exposures, to capture what happens to hues when they are exposed lighter and darker.

Another thought is that if you're not doing it already then you should confirm with a LUT stress test image to ensure you're not accidentally breaking the image.  I use this one and find it very useful: https://truecolor.us/downloads/lut-stress-test-image/

Good luck with this, I've done a lot of camera matching over the last few years and it's a frustrating but interesting challenge and very educational.

1 hour ago, webrunner5 said:

No your just making your life more boring. I grew up with no LUTs. Oh how did we ever make anything look good or even great.  Magic I guess, or maybe a little pre thought before shooting. Like lighting, makeup on and on, but sure things have changed but I think for the worse, a bit of ahh hell just fix it in post mindset. Comes off as a bit of laziness me thinks. Lack of imagination to boot.

 Not really picking on you. I can see you have put some time into doing this. I just am beginning to see all the cameras looking alike and I think NLE's have a lot to do with it to be honest. Not liking the trend.  Blackmagic cameras seem to be the only ones I like anymore. They have bucked the trend.

Actually, doing this is a spectacular way to get a deeper understanding of colour science.  I've matched cameras on multiple occasions and it's an amazingly difficult technical exercise if you want to get a good match, and I've always learned a lot each time I've done it.

LUTs have a poor reputation which is mostly undeserved.  Yes, there are lots of YouTube camera bros out there selling LUTs as the answer to getting good colour, but someone writing a bad book doesn't mean that literature is worthless.  

Apart from the fact that manufacturers supply technical LUTs to take their various LOG formats to a 709 colour space, the use of film-emulation LUTs is one of the best-kept secrets of the professional colour grading industry.  I've heard that the majority of TV shows and movies include the use of a film-emulation LUT of some kind.  It makes sense - if you're a professional colourist you're looking to get the best results in the shortest possible time you'll use whatever tools can do that.  It doesn't mean that those using a LUT aren't knowledgeable enough to have created it, some in-fact did create their own film-emulation LUTs for their own use and their own secret sauce.  Some even custom-wrote their own plugins in scripting language with all the mathematics involved.  Colourists save their own power-grades to apply quickly, and a LUT is simply a type of power-grade.

Using a LUT doesn't mean that the skill of a colourist isn't required.  It's the same as everything else, an Alexa and Master Primes doesn't make you a great cinematographer, using Sky Panels or haze doesn't make your film look wonderful, and using an NLE doesn't mean that you don't have to know how to edit.

Link to comment
Share on other sites

I am also currently down the rabbit hole of trying to find a consistent end result to log that I like.

This level of science is way beyond anything I have tried which has ended up as being FilmConvert Nitrate.

It’s giving me the look I was hoping for.

Still playing with fine-tuning a baseline that I can apply to all my work because yes, I do want a consistent look across all my work.

With your examples, the comparisons of the girl, as with most of these side by sides, my pick is always, “somewhere between the two”.

The Panny one is a touch magenta and yours a touch yellow/green/washed out.

As above, somewhere in the middle would work best for me.

And this is one of the things I like about Nitrate and that is you could just use the ‘film stocks’ just as they are, but you are encouraged to play.

The problem I always find is ALWAYS skin tones.

Without resorting to tweaking the skin tones of every single clip, there isn’t much with an ‘aesthetic’ that keeps them as you’d want without skewing everything else.

That again is why something ‘in between’ can work well, a happy compromise.

Or just tweak the skin tones in every damn clip…

Well done with the effort made and even if it does not end up producing the end result you are looking for, you will have learnt so much from the process.

Link to comment
Share on other sites

So... back to Portra 400. I've been shooting some film lately and although I haven't shot any Portra yet, I've been looking at hundreds/thousands of photos on Flickr and it looks pretty close to me. I like how your LUTS are handling different types of skintones, which is evident in the Caucasian male image and the Asian male and female.

If only you made a set for ML Raw. 

Link to comment
Share on other sites

14 minutes ago, kye said:

I think the colour charts can only get you so far with skintones as they don't have enough hues to work with.  Assuming you're shooting your own film and can point it at whatever you want to, I'd recommend making your own colour checker that has a much more detailed set of hues around the skin-tone hues

Thanks for the suggestion, but it seems that there are already more than enough skin tone hues on the IT8 chart I'm using, from the point of view that I'm not able to match them all 100% accurately without breaking the image. So having more patches wouldn't necessarily help by any means. A broad problem seems to be that of metamerism: patches that have different spectral frequencies that are erroneously understood by the recording medium to be the same colour, introducing erros. I have already experimented with a range of commercially available charts, as well as attempting to design and digitally print my own. But I've found this one to give the best results.

17 minutes ago, kye said:

shoot the charts in a number of exposures, to capture what happens to hues when they are exposed lighter and darker.

Yes, this is the plan. So for Portra 400 and Fuji 400H you can get a usable exposure, arguably, from -3 to +7. The James Miller style LUTs that were so popular in the last decade and sold a million cups of coffee and other hipster products were very much based on the idea of underexposed film: lifted blacks, muted colours and strong colour casts highlight and shadows. On the other hand, slightly overexposed has been perhaps a bit more in vogue in recent years, in the style of Jose Villa's overexposed Fuji 400H wedding photography.

22 minutes ago, kye said:

you should confirm with a LUT stress test image

I have something similar that I've been using but will totally check this out. I can live with a little bit of artifacting, perhaps. And in this version of LUT I've posted, there may well be some: the reds might be a bit funky in places when they get saturated. As I've suggested it's totally a trade off between accuracy and image integrity. It's possible to get a 100% match for all patches but the image is destroyed by this. The trick is to hold on to as much accuracy as possible while avoiding banding etc.

23 minutes ago, MrSMW said:

The Panny one is a touch magenta and yours a touch yellow/green/washed out.

Yes, but I wouldn't necessarily base an opinion on the colour hue on just one image. It's possible that the ambient light in the room was a bit green or that the WB tint on the cam was off - usually the Panny colours are screaming way too much magenta and in this example look quite reasonable. That said, Portra 400 does skew towards green, especially compared to Fuji 400H which is very much magenta, possibly reflecting the differing tastes in desirable skintones between the American (Kodak) and Japanese (Fuji) markets. You're right about the saturation though. I find it slightly low. Now, Portra is supposed to be relatively low saturation. But bear in mind that what I'm emulating is a specific roll of film, that has degraded since it left the factory according to the conditions it was stored in, that has been developed in a specific chemistry that may deviate more or less from the ideal C-41 recipe. This is a long-winded way of saying that I have seen slightly different results from different rolls of the same film stock shot at different times and developed in different places. It's something I'm going to look into a bit more.

30 minutes ago, MrSMW said:

Or just tweak the skin tones in every damn clip…

With my lut it should be sufficient to do a little bit of white balance tweaking. But I've observed that it's, for example, sometimes towards red and sometimes towards green as you would expect with normal variance of colour temperatures. If you try to globally WB the Panny colours for skin then everything goes green.

 

8 minutes ago, mercer said:

I like how your LUTS are handling different types of skintones, which is evident in the Caucasian male image and the Asian male and female.

Absolutely! And I'd guess that what Portra does is push greenish skin and magenta skin more towards the same hue. That's why skin tends to look so uniform with it. Definitely give Portra a go! It's expensive but the results are slightly nicer than the cheaper film stocks.

10 minutes ago, mercer said:

If only you made a set for ML Raw. 

At some stage I will make a set for Camera RAW and Lightroom for digital RAW, but that may not be the preferred workflow these days.

Link to comment
Share on other sites

3 minutes ago, hyalinejim said:

Yes, but I wouldn't necessarily base an opinion on the colour hue on just one image.

Yes I agree.

The 'problem' with judging colour science with screen grabs or photography, especially in side by side comparisons, is that your brain doesn't have time to adjust.

If it's an entire movie or short film and has a specific colour, tone or whatever, we accept it much more easily, even if the colours are actually way off in regard to reality.

This is especially evident in sci fi stuff like; The Martian, Mad Max, Dune etc where none of the colour in the entire move it exactly 'real'...but it becomes 'real' to us very quickly.

I've looked at a few wedding films of a couple of folks I really rate and watching those, everything looks Captain Amazeballs.

But when I have done some screen grabs and looked at those as individual pictures, I like it a lot less.

We need to look at these things in isolation IMO, ie, the entire film needs to gel, same as a set of images needs to do the same.

That's what I am working on myself, - a film grading style that is not too heavy handed, but is mine and ditto a photography grade that is similar with both complimenting each other.

Link to comment
Share on other sites

12 minutes ago, hyalinejim said:

Thanks for the suggestion, but it seems that there are already more than enough skin tone hues on the IT8 chart I'm using, from the point of view that I'm not able to match them all 100% accurately without breaking the image.

I think this is where the art and true deep-knowledge come into this whole challenge.

I don't know what tools or techniques you're using to make your adjustments, but there are a number of ways to accomplish something, and certain ways are more or less likely to break an image.

I highly recommend watching Juan Melaras YouTube videos where he matches things (and if you can find it, he did a video replicating the Linny LUT but pulled it down - maybe it's available somewhere else).  I suggest this because he does a bunch of really cool things using alternate colour spaces (HSL, YUV, Lab, and more) and using tools that won't break the image such as the channel mixer or curves.

It's obvious that Juan can look at the response of a LUT or look and see the big picture and knows what global tools can align to that look the easiest way without breaking it.

Have you read a lot of resources about film emulation?  Happy to share my bookmarks on it if you're interested.

12 minutes ago, hyalinejim said:

A broad problem seems to be that of metamerism: patches that have different spectral frequencies that are erroneously understood by the recording medium to be the same colour, introducing erros.

What reference setup are you using?

I would imagine that you'd want to use as high a quality of light you can (definitely a black-body radiation source) and either the sun or a halogen lamp, and then just ignore all other lights as they will be inferior.

The other challenge will be the shape of the RGB spectrum sensitivities, eg:

?ACT=44&fid=17&d=9202&f=provia_velvia_se

The way that Juan has matched these is with the RGB mixer, although there might be some situations where that won't be completely effective, not sure.

12 minutes ago, hyalinejim said:

But bear in mind that what I'm emulating is a specific roll of film, that has degraded since it left the factory according to the conditions it was stored in, that has been developed in a specific chemistry that may deviate more or less from the ideal C-41 recipe. This is a long-winded way of saying that I have seen slightly different results from different rolls of the same film stock shot at different times and developed in different places. It's something I'm going to look into a bit more.

Yeah, huge variation exists between batch, processing method and labs...

Steve Yedlin had a lot to say about that in this article that I'm assuming you're familiar with: http://www.yedlin.net/OnColorScience/

A phrase for colour matching that I really like is "in the same universe".  It accounts for how closely you need to match individual shots in an edit, but it also accounts for what @MrSMW says about getting used to the look while watching a film, which is absolutely a factor too.  We never watch the same scene through two different cameras / processes / grades, so differences have quite a degree of tolerance.

12 minutes ago, hyalinejim said:

With my lut it should be sufficient to do a little bit of white balance tweaking. But I've observed that it's, for example, sometimes towards red and sometimes towards green as you would expect with normal variance of colour temperatures. If you try to globally WB the Panny colours for skin then everything goes green.

I have a theory that making a grade that allows for WB adjustments should be completely possible.  I never got around to trying it, but my theory is this.

The camera does things in a certain order:

  • light comes in and hits the sensor with its spectral sensitivity
  • the camera applies the WB
  • the camera applies the colour profile

I think that the secret is to organise your adjustment so that it peels-the-onion by reversing the order of operations, like:

  • undo the colour profile (GH5)
  • undo the WB (GH5)
  • adjust the spectral sensitivity from the source camera (GH5) to target (film stock)
  • apply WB adjustment of target (film stock)
  • apply the colour properties of the target (film stock)

The challenge is to separate the three layers, which in my case I was matching two digital cameras and so you could just take a RAW still image in each which allows you to separate out the colour profile from the sensor and WB.

I think you could potentially still apply some of this logic by building the adjustment in a modular way and shooting the colour checkers in a range of WB situations (maybe using gels?).  Then you might be able to adjust the WB and spectral sensitivity adjustments in their own nodes and you could see if they are compatible with the same colour profile in matching the reference images.

The order of operations isn't completely clear in my head, but I think you need to do controlled tests to work out the spectral sensitivities first using a proper WB, then the WB adjustment by shooting RAW stills in different WB settings, then the colour profile.

Hopefully that made sense?  It would require getting the order of operations completely down, and executing in a meticulous way, but if I'm right then it should be able to be done.

12 minutes ago, hyalinejim said:

At some stage I will make a set for Camera RAW and Lightroom for digital RAW, but that may not be the preferred workflow these days.

If you make it modular like I suggest then once you've done the GH5 version the Canon RAW version would be super easy.

The adjustment would only require reverse engineering the spectral sensitivity and WB on the Canon and then applying the properties of the film stock which you've already done.

Juan Melaras discussion on how he made his BMPCC 4K and 6K to Alexa conversions, and in my GH5 to BMPCC thread were very interesting and I think because he was just going from RAW to RAW he didn't have to nullify colour profiles in either camera, just having to do the two lower-levels.

Link to comment
Share on other sites

Great write up and presentation. Thanks. Back then you matched the Gh5 to Canon 5d Raw. Now mercer wants you to match the 5d back to the Gh5.😊 @hyalinejim
Little joke, dear friends! Would also love some 5d3 magic in Kodak and Fuji flavors!

I am looking forward to test this Lut. Until now I still prefer grading the Gh5 more than the S1. Could be that Gh5 is supposed to be in 709 color space when in VLog while the S1 is in VGamut. The larger color space giving a different response and metric when working curves. Anyway, thanks a lot and looking forward to what we can achieve with your lut.😊

Link to comment
Share on other sites

14 minutes ago, kye said:

I would imagine that you'd want to use as high a quality of light you can (definitely a black-body radiation source) and either the sun or a halogen lamp

Yes, but metamers can still apparently occur even with a high quality light source 😞 In the chart I use I see this a lot with reds, where two similar-ish reds are fighting each other in the LUT. One wants to be bright and saturated and the other wants to be dark and subdued. If I let them both have their way then it breaks the LUT. So there may be some sort of mix-up happening between the actual spectral data of the patches as they are in reality, and how that's recorded (and scanned) on film and on digital acquisition.

15 minutes ago, kye said:

I have a theory that making a grade that allows for WB adjustments should be completely possible

If you can get your footage into ACES CC or CCt colour space (and you can do this in Resolve or After Effects) then a simple shift of the RGB channels will do a very good job of white balancing, almost as if the footage was RAW. I usually leave my GH5 on Daylight WB and then do this in post. But the footage probably needs to be log and at least 10bit. What do you shoot on the GH5 if you're not shooting log?

10 minutes ago, PannySVHS said:

I am looking forward to test this Lut

Great! Let me know what you think about the contrast and saturation and colour in general 🙂

11 minutes ago, PannySVHS said:

Could be that Gh5 is supposed to be in 709 color space when in VLog while the S1 is in VGamut

It looks to me like GH5 V-Log is not in Rec709 colour space but certainly lots of people agree that the Panasonic V-Gamut to 709 conversion does not work all that well for GH5. I've tried some LUTs that I built on the GH5 on the S1 and they looked fine to me without making a direct comparison of the same scene shot on buth cameras, but I guess the accuracy of some hues will be a little off.

Link to comment
Share on other sites

53 minutes ago, hyalinejim said:

Yes, but metamers can still apparently occur even with a high quality light source 😞 In the chart I use I see this a lot with reds, where two similar-ish reds are fighting each other in the LUT. One wants to be bright and saturated and the other wants to be dark and subdued. If I let them both have their way then it breaks the LUT. So there may be some sort of mix-up happening between the actual spectral data of the patches as they are in reality, and how that's recorded (and scanned) on film and on digital acquisition.

I suppose that it could happen - it's easily possible to build a LUT where multiple input values go to the same output variable, in which case reversing that adjustment would be impossible, although I'm guessing that it's not likely.  I would suggest that the Panasonic colour science may be compressing a certain part of the colour space (which is flattering to pull skintones closer together so is definitely a thing that happens) but I would doubt they'd completely crush anything.

That's why I was suggesting a higher resolution colour chart - it might be that two patches are very similar but if you knew what skewing was happening around those then you might be able to see what was happening and find a way around it.  I did a bunch of tests in Resolve by using two 10-step generators (one horizontal and one vertical) to create a 10x10 grid of things like Hue vs Saturation etc, and pointed the camera at the monitor in a dark room.  You can "zoom in" to the hues and see what the GH5 is doing.  That's relatively easy and might be a quick exercise to get a feel for those trouble spots.  

You could even setup the computer with one of those, film it with the GH5, put the GH5 feed into a monitor that applies the LUT and displays a Vectorscope, and then in Resolve you could play with the Hue and Sat controls in the LGG panel to move the grid around, and watch what happens on the Vectroscope.  If it pinches or ripples or something then it should be pretty obvious.

I wouldn't suggest that a monitor is a good source of light, but it might be useful to get a sense of what the colour profile of the GH5 is doing.

53 minutes ago, hyalinejim said:

If you can get your footage into ACES CC or CCt colour space (and you can do this in Resolve or After Effects) then a simple shift of the RGB channels will do a very good job of white balancing, almost as if the footage was RAW. I usually leave my GH5 on Daylight WB and then do this in post. But the footage probably needs to be log and at least 10bit. What do you shoot on the GH5 if you're not shooting log?

I'd either shoot HLG, which isn't quite rec2100 or rec2020 but is close enough, or I'd shoot CineD.  The challenge is that I shoot in 24p, 60p, and 120p and the HLG profile isn't available in 120p mode, so for a while I was shooting in CineD on all of them so I'd have the same colour in post.  Since then I've changed back to HLG and just CineD in 120p because I can match things easily enough in post and I appreciate the extra DR.

I also shoot the same projects (trips) on the Sony X3000 action camera (prior to that it was a GoPro), and my current iPhone of the time.  There aren't ACES profiles for either of them, so there was no advantage to the ACES compatibility of the Vlog upgrade for the GH5.

I kind of went down the colour grading rabbit hole and ended up buying the BMMCC to copy the colour science on the GH5 (as I couldn't afford an Alexa to compare it with) but since then I've learned a lot about colour grading and how to get good colour and what that means so I'm now less concerned with that detail.  Since then Resolve implemented the Colour Managed feature which does a great job of adjusting WB etc in post with the rec2100 colour space.  I also worked out how to get nice results when grading under a film emulation LUT (I like the 2393 LUT quite a bit).  Now I'm more focused on cinematography and editing, as I can get colour that's good enough for my purposes.  I'm not winning any awards, that's for sure, but I still wouldn't be if I managed to perfectly replicate one camera with another.  

On the colourist forums there is thread after thread of people getting super detailed about tiny little details to do with film-emulation, and then in the other threads there are casual comments that mention that a PFE LUT is one of the 20+ adjustments they typically do to make a look, or that in real projects they can get away with one or two nodes instead of a PFE to give a bit of a film vibe because the rest of the look is obscured by the other adjustments.  The number of times when an authentic film look is really required seems to be pretty low, and most people get a good enough 8mm simulation if you just blur, add gate weave, apply a simple HueVsHue and HueVsSat, and apply a ton of grain.  I've seen threads where people have said that they've never before seen a film grain emulation that looked remotely real, and that they worked with film for 30 years so they are especially attuned to it, but then you've got Walter Volpatto (about as senior a colourist as you get) saying that he doesn't even add grain on most projects anymore because the streaming compression kills it.

Link to comment
Share on other sites

OK, had a play with your LUT plus a couple of FilmConvert alternatives:

Premiere, S1H log footage

The order is:

1 = The Portra Lut as provided, no tweaks

2 = The Portra Lut tweaked...albeit, I brought the highlights down too far 🤪

3 = Nitrate's version of Portra based on S1H log settings no tweaks. Err, a bit purple innit?! I wouldn't call that very 'Portra'!

4 = Nitrate Astia no tweaks

5 = Nitrate Astia tweaked

6 = Nitrate Velvia tweaked

7 = Nitrate Astia best I could get it 'globally' ie, without any masking skin type tweaks etc + with S 35mm grain

8 = As 7, but + FF 35mm grain

(7 = the closest I have got to obtaining the look I am after for my own work...but is still a work in progress)821885639_APortraNoEdit.thumb.png.ec284693bf80eff40821722c48551bfe.png1111590719_BPortraTweaked.thumb.png.98b19d971b8977e94d8d55ec957baa3e.png1754300721_CNitratePortra.thumb.png.5f4446071eeb099b592291c4b66cc24a.png866439207_DAstiaNoEdit.thumb.png.ee78553edda04e9f5f50bc30e6257f1a.png1082203075_EAstiaEdited.thumb.png.5210e4a9d4f9980e0d37ffe6c7a0a7af.png1764139658_FBestSkin.thumb.png.3bef77664abba8132b9004c28c7b8a11.pngG.thumb.png.69699c2c2b2abdeb3d69ee53466c6dc0.pngG2.thumb.png.252a6c91d300da8b1abc1433e8f6a459.png

 

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

 Share

  • EOSHD Pro Color 5 for All Sony cameras
    EOSHD C-LOG and Film Profiles for All Canon DSLRs
    EOSHD Dynamic Range Enhancer for H.264/H.265
×
×
  • Create New...