Best skin tones

[Ryan De Franco]

Adrian, valid points all around. But what do you do when, on the camera monitor, the skin tones look plastic;

at the DITs OLED monitor, the skin tones look plastic,
after a redlogfilm first light, the skin tones look plastic
in the grading suite, you spend an hour with the colorist, but skin tones still look plastic
no matter where you watch the final product--on a laptop, on a digital projector, hell on a film-out.... skin tones look plastic...

is it still the projector's fault?

 

[Adrian Jebef]

Adrian, valid points all around. But what do you do when, on the camera monitor, the skin tones look plastic;

at the DITs OLED monitor, the skin tones look plastic,
after a redlogfilm first light, the skin tones look plastic
in the grading suite, you spend an hour with the colorist, but skin tones still look plastic
no matter where you watch the final product--on a laptop, on a digital projector, hell on a film-out.... skin tones look plastic...

is it still the projector's fault?

Ah, yes. Then this is when you employ a good colorist at $1500/hour and let him do multiple skin tone passes. Separation, multiple color-correction layers, blurs, comps, sat/luma/hue compensations, etc.

I agree that digital skin tones (from either F65, ALEXA, or EPIC) can be quite lacking when compared with a film projection. That being said, it is a digital world. I've found the digital projection (SONY 4K) from features shot on film exhibits these same limitations. Unfortunately there is no going back. Again it's really all about final display in the end, not necessarily the point of acquisition. Kind of destroys the need for film when it's just going to be scanned, digitized, sharpened, and made to look like video in the end...

 

[Ryan De Franco]

Why will it be made to look like video? Look at the Red films I posted earlier, look at The Master (which had no DI) or Samsara (which had one, shot 65mm, but looked great on the Christie 4K projector at Toronto LightBox). Be it film or digital capture, none of these "look like video in the end." Why should we assume they should?

The trouble with $1500/hr for "multiple skin tone passes" is that such an approach works great for commercial, TV, promo, etc--where talent stays in one nicely lit spot, and the result can look "processed and clean." In narrative, where light can and should change on an actor's face--and where colorist slush money is more limited--what are we supposed to do? Furiously grade every shot of every actor, spending a $1500/hr every time the light changes, and end up with something that looks like a polished commercial?

By then you have far exceeded the costs of originating on film. And then what is the point? To have a camera my assistant can warm her hands with in winter?

Please Red, make the new sensor and its color science look damn good on human beings. That's all we ask.

 

[Evin Grant]

I have to wonder if some of this is a lack of texture IE grain as much as color accuracy. Grain has a dithering effect that helps to break up tonal transitions and it may be that adding a bit of organic texture back into the image would help a bit. It's not an end all be all solution but it may help.

 

[Rob Castiglione]

That is interesting speculation about grain and its dithering effect - just like audio to mask the steppiness of digital audio signals. There seems to be at least some partial analogy between the introduction of high end digital audio and what is now happening to digital images.

 

[Ryan De Franco]

Evin and Rob, that's a really, really good point. I haven't experimented with adding grain to digitally captured motion--it feels like using a blue streak filter on spherical lenses, really, who are you kidding--but I think I need to try it. Thank you for that advice.

Skin "tone" is as much about tonal transition as it is about "color."

 

[Ron Smith]

Evin and Rob, that's a really, really good point. I haven't experimented with adding grain to digitally captured motion--it feels like using a blue streak filter on spherical lenses, really, who are you kidding--but I think I need to try it. Thank you for that advice.

Skin "tone" is as much about tonal transition as it is about "color."

You may be on to something. If you look closely at the Ruby Sparks trailer in 1080p or 720p, it is quite apparent that grain was added to the image. That may contribute to the natural look of the movie.

 

[Juan Melara]

Adding grain goes a long way in removing the digital look by adding subtle texture. It also seems to increase the perceived sharpness of the image. Read DI Colourist Adrian Hauser's post on this subject here.

Below is a sample image that I've applied grain too. The original is a Scarlet shot by Will Wernick, find the R3D in this thread here. I've also included a with and without grain comparison.

Graded in Resolve. Grain added in photoshop. Actually, I probably shouldn't call it grain since it's just a 2.2% Guassian Noise filter! But it's late and I couldn't be bothered firing up After Effects etc, but you get the idea...

Nothing overly fancy going on in Resolve, I just take care in avoiding grading decisions that end up creating plastic skintones. Pretty simple.

 

[David Battistella]

There are two very important considerations here that I think may have gone unmentioned. First, skin tone, color, quality, sheen, luminosity, transparency of the human face has a very wide ranging variation. Pantone has over 100 swatches of skin tone now and the old vector scope has exactly ONE line shooting out from the center to indicate a ballpark skintone.

http://www.designboom.com/art/pantone-skin-color-spectrum/

The rendering of skin and skintone in digital and filmed mediums depends much more on lighting and makeup and research into these two things, than it does on a particular brand of camera.

i see it this way. The epic is a non discriminating digital capture device where the manufacturers have gone toward what I call; a true color experience. The out of the box camera color spaces give a very accurate reproduction of a scene, including skin tone when the images are processed properly. It is also an extremely high definition image.

Other manufactures do more to the image or mimic what we "perceive through a lifetime of watching" as correct.

What I am saying is this. We have been exposed to a certain way of seeing skin that we deem as correct when in fact it is not technically correct it has been enhanced. So when considering skin tones with RED you must understand that it is reproducing color more accurately than ever in history.

When people complain of bad skin tones in with red it's because they are seeing more than they are used too, they are seeing lighting and makeup techniques that apply to a different medium, one other manufacturers have tried to emulate (which doesn't mean it is correct, it's just correct to you) and they are sticking with what people are used to seeing.

with the range of skin tones we now ask cameras to reproduce and the massive jump in resolution, mostly what takes getting used to is seeing an actual correct skin tone and knowing that you have to compensate and perhaps change your methods to achieve what you are "used to".

this whole skin tone argument is the brain seeing through this new medium stuff it has never really had to process. Without thinking this through and applying techinque (the artists job) people tend to blame the tool. It's not the tool. The tool is providing reference images. It's up to the individual artists to determine what Is correct. That is where the standard lies.

David

 

[Bruce Allen]

What I am saying is this. We have been exposed to a certain way of seeing skin that we deem as correct when in fact it is not technically correct it has been enhanced. So when considering skin tones with RED you must understand that it is reproducing color more accurately than ever in history.

Really? Have you done objective tests or is this just an assertion?

Personally I think RED's color accuracy is currently poor compared to, say, a good DSLR still. I suspect that the color filters are not particularly pure.

Especially when you look at red tones, which seem to go either pink or brick color way too easily.

The only objective tests I've seen done on this are Adam Wilt's for provideocoalition - which shows a bit of weird crosstalk between channels.

If you have some tests that objectively show RED to have superior color accuracy, I'd love to see them.

Bruce Allen
www.boacinema.com

 

[Bruce Allen]

Nice! That's really good for RED skin tones!

Yeah, I am a big fan of adding grain too. It definitely improves perception of detail and jives well with human vision.

Actually I do a little NR using Neat Video first - which removes some of the Wavelet compression artifacts. Then add grain on top. CineGrain's 4K scans stuck on top in overlay mode work pretty well for me although I understand that a more accurate film look would distribute grain more non-linearly.

Talking of Wavelet compression... ouch! It is really visible in this clip. Maybe the compression is contributing to the plastic skin feeling in RED footage.

Cheers man - nice one.

Bruce Allen
www.boacinema.com

 

[David Battistella]

Bruce,

I'm thinking more along the lines of filmed images than the DSLR since we are dealing with a cinema camera. I'm reflecting on what I am noticing based on images I am seeing from the camera compared to film and the overall point is that most of us (most) have been looking at lower resolution images where skin is washed over and that there are many factors that go into determining a "good" skintone. DSLR's capture one frame and they do that remarkably well but they do not do what RED does.

I take your point and my post is subjective, it's not science.

I think we can both agree that the color accuracy is has improved dramatically and is on par with film stocks I have seen (which have their own characteristics)

David

 

[Brandon J.F.]

Maybe the compression is contributing to the plastic skin feeling in RED footage.

That could be. I shot a live performance at 10:1 and noticed the singer's face looked a little "photoshopped" without any grading done at all. Here's a graded still:

Where in your workflow do you use Neat Video NR?

 

[Bruce Allen]

That could be. I shot a live performance at 10:1 and noticed the singer's face looked a little "photoshopped" without any grading done at all. Here's a graded still:

Difficult to tell... can't see what's going on there too well. What does it look like if you decrease exposure and contrast, decrease the amount of green a little and increase saturation quite a bit?

Also, try various levels of sharpening and diffusion. Those are not mutually exclusive - it's just about choosing what frequencies in the image to emphasize or diffuse. Also you can choose to emphasize chroma or luma....

Where in your workflow do you use Neat Video NR?

Currently: good old After Effects for Neat Video. Switching to Nuke though for multiple reasons, mostly speed, workflow, EXR performance, pipeline-ability...

I like to do preliminary color correction in Resolve without NR. Then if I hit issues, add NR to taste using After Effects to the source footage before color correction. You don't want to add too much... it's very delicate! Personally I use really low settings, but turn the temporal noise reduction on - that cleans up compression artifacts better for me because it's using the info from multiple frames but in a non-heavy-handed manner.

The I replace the footage in Resolve with the NR'ed footage. Sometimes NR helps a lot with getting cleaner secondary keys. NR is part of a general pre-processing stage which can also include chromatic abberation removal (you'd be surprised how much it can screw up your color correction keys!).

I used to be religious about staying RAW - now I've gotten tired of that and just try to get white balance and exposure vaguely right, then go to RedLogFilm and DPX files. If it's good enough for the guys grading Prometheus, it's good enough for me! I chatted to the guy grading it at the RED event and he had no clue what R3D development settings were used... just get the DPX files from the assistant and go! And hey, 10 bit log seems to be able to contain all of the beautiful colors and latitude of film pretty well... Of course I do a lot of stuff with VFX so that makes sense for me. If you're not doing a lot of VFX things then maybe it makes more sense to stay RAW until longer.

I'm experimenting with EXR files and linear workflow but ACES is a little unreliable presently. At the moment, I'm actually using Log workflow... but storing everything within EXR files because it has lossless compression. Bizarre and not recommended but fun Working in lightly-compressed 10-bit Log DNxHD or ProRes is next on my list of things to try for a project...

Here's my old workflow - it's a little out of date:
http://reduser.net/forum/showthread.php?68295-Color-Correction-for-quot-TimeScapes-4K-quot

Bruce Allen
www.boacinema.com

 

[Bruce Allen]

Bruce,

I'm thinking more along the lines of filmed images than the DSLR since we are dealing with a cinema camera. I'm reflecting on what I am noticing based on images I am seeing from the camera compared to film and the overall point is that most of us (most) have been looking at lower resolution images where skin is washed over and that there are many factors that go into determining a "good" skintone. DSLR's capture one frame and they do that remarkably well but they do not do what RED does.

I take your point and my post is subjective, it's not science.

I think we can both agree that the color accuracy is has improved dramatically and is on par with film stocks I have seen (which have their own characteristics)

David

Thanks for the nice discussion, David!

Totally agree... yes, you hit the nail on the head: I would indeed be happy if, given variable lighting conditions, RED camera footage always just looked like a series of great stills from a DSLR. That to me would really challenge film, since DSLR colors can either beat or come close to film in many conditions in terms of capturing the beautiful details I'm in love with.

Also agree that RED may be more accurate overall currently. I just think that RED must be missing some details... if RED were capturing everything necessary it would be easy to distort RED colors to match film. But it's so damn hard!

Bruce Allen
www.boacinema.com

 

[David Battistella]

.. if RED were capturing everything necessary it would be easy to distort RED colors to match film. But it's so damn hard!

Bruce Allen
www.boacinema.com

I hear ya Bruce, but we know that in order to keep the data sizes manageable redcode has to exist. It's a new beast' an evolving film stock in the hands of many who are both applying what they know and experimenting with how it can be pushed.

I see how they have played with the blue channel and how adjustments move the image around, but that's the fun part right? All of us figuring out how to squeeze the best images out.

Its not film, and that is ok. It's not tape (and that is better) but its nice to be in the garage and under the hood as it evolves.

The whole thinking on what skin tone is needs to evolve too because of digital image capture.

David

 

[Mike S. M.]

Ultracon filters and the like do really nice things for humans.

In my experience using Ultracons in situations not stressed by high contrasts makes for very dull images.

 

[Björn Benckert]

That could be. I shot a live performance at 10:1 and noticed the singer's face looked a little "photoshopped" without any grading done at all. Here's a graded still:

Where in your workflow do you use Neat Video NR?

Red gamma3 does cut the highlights a bit. So before going LIN i would make sure to add a soft clip. If you have this singer as a r3d then I might be able to show you something interesting involving kodak print 3D luts...

 

[Brandon J.F.]

Red gamma3 does cut the highlights a bit. So before going LIN i would make sure to add a soft clip. If you have this singer as a r3d then I might be able to show you something interesting involving kodak print 3D luts...

Very cool Björn. I'll find the r3d.

 

[Jared Knecht]

Has anyone used Film Convert? I am debating about purchasing the plug-in. Really interesting execution not just a 'filter'. Skin tones look great! It would be interesting to use it as a quick starting point before furthering correcting and grading.