a7s skin tones: slog2 vs pp off (shogun 4k)
- Thread starter caseywilsondp
- Start date
caseywilsondp
Well-known member
It's all very subjective. I think many people went with the stylized look, punching orange and reds. A few focused on natural skin tones.
Obviously I'm partial to the non slog as that's how we lit the scene and captured what we were after. That being said, there is a technique in adobe speed grade where I can input the neutral image we lit with and apply the look onto the slog footage. This is beneficial because slog still retains more information so I can pull back the filaments in the lightbulbs, and pull up the shadows and details on her dress, and still keep that amber/red look that permeates the neutral picture.
Slog2 isn't easy to grade by any means, and the look ultimately is up to the colorists taste. I'll upload my finished look soon to show what I finally settled on.
nyvz
Veteran
It's probably worth noting that the framing is a bit different on the SLOG and the PP-off version. I like the wider framing of the PP-off frame a bit more since it keeps in the nice looking practicals better. Frankly I don't disagree that the PP-off looks better than a lot of the grades people have posted, but SLOG2 does come with some added complexity which is the trade-off for benefits that may be seen in certain situations. It's odd to me that more people aren't using LUTs, as I think to some degree a LUT is mandatory for SLOG2 since it's not that easy to correct S-gamut by hand... SLOG2 can be more complicated to expose and whitebalance and grade than a WYSIWYG gamma curve but if you do and know what you are doing it can go a long way for scenes where you need the flexibility to adjust exposure and over/under DR in post and if you want to use great film emulation LUTs. I think a lot of people overcomplicate SLOG2 by trying to grade it even though they don't fully understand it, and that's where LUTs come in, there is no reason to reinvent the wheel.
I always shoot SLOG2 when it's up to me and have been very happy with it, but I have a healthy collection of LUTs to make it quite simple when necessary.
I always shoot SLOG2 when it's up to me and have been very happy with it, but I have a healthy collection of LUTs to make it quite simple when necessary.
caseywilsondp
Well-known member
What're your go-to LUTS to get a standard image out of slog2?
caseywilsondp
Well-known member
is the sony lut you speak of from their f55?
Casey's PP Off on left, Slog2+SGamut corrected to look natural on right
A couple thoughts: color perception changes with viewing scale (shrunk my examples down), and due to the non-linear nature of Slog2, parts of the image 'under exposing' turn magenta, and 'over exposing' turn green (as shown in the video Samuel posted). The magenta/green shift with exposure may be minimized if using a SGamut => rec709 LUT. However, note in the close up right image, the nails are no longer red and are now pinkish (extra blue needed to get skin looking closer to normal). Makes a case for lighting for WYSIWYG and using a very color accurate profile (little or no post color needed), or perhaps shoot in normal light (at least for the face/skin) and do mood coloring in post.
Using reference images is helpful: taking shots with the 5D3 (RAW or stills) provides a very helpful reference for accurate color. Resolve's color chart Color Match feature doesn't really work: better to use a 5D3 (or Nikon, or ARRI
) reference.
The challenge with this shot: the lighting is close to the model's skin color and hair: pink/red/amber/orange. The background and the model blend together. To get more color separation, the model needs to be more blue (complement of yellow): second wide set shows more blue to separate model from background.
When trying to make the skin tones look pleasing, many of the examples ended up looking sepia- not much color separation, almost monochromatic. That's a viable look if that's the goal. I'm a Technicolor 3 strip fan- love intensely saturated and separated colors when high energy or excitement is the goal. The intense amber (red & green) light with low blue makes the resulting image 'thin'- very little leeway in post, and thus a bit tricky to grade. To my perception, the 'blue' version pops and draws my attention the most.
A couple thoughts: color perception changes with viewing scale (shrunk my examples down), and due to the non-linear nature of Slog2, parts of the image 'under exposing' turn magenta, and 'over exposing' turn green (as shown in the video Samuel posted). The magenta/green shift with exposure may be minimized if using a SGamut => rec709 LUT. However, note in the close up right image, the nails are no longer red and are now pinkish (extra blue needed to get skin looking closer to normal). Makes a case for lighting for WYSIWYG and using a very color accurate profile (little or no post color needed), or perhaps shoot in normal light (at least for the face/skin) and do mood coloring in post.
Using reference images is helpful: taking shots with the 5D3 (RAW or stills) provides a very helpful reference for accurate color. Resolve's color chart Color Match feature doesn't really work: better to use a 5D3 (or Nikon, or ARRI
) reference.The challenge with this shot: the lighting is close to the model's skin color and hair: pink/red/amber/orange. The background and the model blend together. To get more color separation, the model needs to be more blue (complement of yellow): second wide set shows more blue to separate model from background.
When trying to make the skin tones look pleasing, many of the examples ended up looking sepia- not much color separation, almost monochromatic. That's a viable look if that's the goal. I'm a Technicolor 3 strip fan- love intensely saturated and separated colors when high energy or excitement is the goal. The intense amber (red & green) light with low blue makes the resulting image 'thin'- very little leeway in post, and thus a bit tricky to grade. To my perception, the 'blue' version pops and draws my attention the most.
nyvz
Veteran
Yes, it's not designed for the A7S but it is designed for Slog2+Sgamut and works pretty well as a good starting point.
Samuel H
Major Contributor
I don't think you're crazy. Some of the slog graded shots are good but that original shot is great too. Getting it right on camera is often what leads to the best results, the problem is that if you bake in an image and get it wrong in camera, you're screwed, so it's safer to shoot with something malleable like slog.
MichaelFischer.at
Active member
Ok, since nobody did it, here's a S-Log/S-Gamut to PP off (Standard) match.
The way I achieved this could fill a separate article, but I did a pseudo-ACES workflow, similar to that one:
http://www.dvxuser.com/V6/showthread.php?333189-ACES-full-circle
Had a little jump start since I recently created a S-Gamut => r709 Matrix/Preset in Resolve.
Seems you almost need an engineering diploma to color correct Sony's S-Gamut footage...
The way I achieved this could fill a separate article, but I did a pseudo-ACES workflow, similar to that one:
http://www.dvxuser.com/V6/showthread.php?333189-ACES-full-circle
Had a little jump start since I recently created a S-Gamut => r709 Matrix/Preset in Resolve.
Seems you almost need an engineering diploma to color correct Sony's S-Gamut footage...
looking very good indeed! Would be very interesting to see your approach on this 'psuedo-ACES' workflow!
cheers/ Pat
I agree. You are not crazy. Slog2 is a nice thing for DR, but I'm yet to see any reason to use sgamut with it (on the a7s anyway).
Samuel H
Major Contributor
paulcurtis
Veteran
You're not really supposed to grade from an SGamut image really. Nor an AlexaLogC or WideGamut for example. And there should be a separation between Gamma and colorspace because as you know you can set the camera to Slog2 gamma but choose a different colorspace. Those Sony LUTS have both the gamma and colorspace baked in and that's not ideal in this case
You want a workflow where you can separate the two.
I can choose to say this footage is in Slog2 gamma (so i have the full 13 stops of range in it) but it is recorded in the SGamut colorspace. The SGamut is an odd space designed by sony to maximise the colours that their cameras can capture. They obviously felt that existing colorspaces weren't good enough. So when you look at SGamut it's not going to look real unless a conversion has happened into a viewing space. Sure you can massage the colours and mash it around to get it looking okay but that was never their intention.
Here's a quick sample just taken in correctly, bit of contrast and i'd rolled my own film print LUTs in Nuke so popped that on at the end. I don't know what the original intent of the grade is, but i know that as a starting place that grade could go anywhere from the starting image.
cheers
Paul
MichaelFischer.at
Active member
Hi Pat, my node structure/order looks like this:
1) S-Log2 => Linear LUT
2) S-Gamut => r709 Matrix via RGB Mixer
3) Exp/WB correction
4) Further color correction via the Hue/Hue and Hue/Sat curves
5) Linear => S-Log2 LUT
6) S-Log2 => r709 1D LUT
7) Soft Knee so that I more or less end up with Sony's 709-800 curve
I call this pseudo-ACES since I'm doing a conversion to linear light, like ACES does it. The difference is, I'm stayingin the r709 space compared to ACES' far wider colour space.
Paul, btw, very nice grading!
remsyatassi
Well-known member
Some of these grades are pretty bad. One or two are nice. Making the model look yellow and contrasty - especially highlighting her blemishes on the closeup - are just odd choices to me. I can't see a client ever going for that kind of look, it's kind of hideous.
nyvz
Veteran
The benefit of S-gamut is that you can ostensibly LUT it and get to REC709, P3, REC2020, film emulation, etc. simply using LUTs that already exist or will soon exist. It is not easy to correct manually, that's partly why in Sony's other cameras they started using Sgamut3.cine, since Sgamut and Sgamut3 both capture wide gamuts that are not quite on axis with REC709, particularly on the green axis. Sgamut3.cine comes with the green axis corrected out of the box to some extent so it scales more easily down to Rec709/P3/REC2020 by hand without special transforms.
Not sure why anyone wouldn't use a LUT as a starting point unless they really have their own special sauce they know works better for them.
Not sure why anyone wouldn't use a LUT as a starting point unless they really have their own special sauce they know works better for them.
caseywilsondp
Well-known member
I'm still trying to figure out how I like to best tackle slog2/sgamut. This time around, I used 'match to master playhead' in speedgrade to let the computer match the slog2 shot to how the pp off shot looked, and it worked remarkably. I'll upload the final grade we went with soon I hope.