Cinestyle vs Neutral vs RAW Test

[Sareesh Sudhakaran]

I have crudely tested the Technicolor Cinestyle picture style for Canon DSLRs for video along with Neutral and RAW here: Technicolor Cinestyle vs Neutral | Sareesh Sudhakaran

My conclusion: Technicolor Cinestyle is only slightly better than Neutral. For HDSLR filmmaking, I'd rather shoot Neutral/Faithful in camera and avoid heavy post processing.

However, RAW beats Cinestyle hands down.

I would love to hear what others have to say about Cinestyle. Does your experience match my own, or is it contradictory?

[Jon Fairhurst]

Nice test!

The Cinestyle shot definitely has more dynamic range than Neutral. You can see detail further into the light as well as further into the shadows.

I definitely recommend Cinestyle when shooting outdoors. On your test shot, the extra range at the top just gives you a couple inches further into the light bulb. On a highly exposed overcast shot, it could make the difference between capturing the sky or having the whole sky blown out.

I agree that shooting with Cinestyle takes more effort in post. I just apply an RGB curve to the whole track (or a single clip as needed) in post, rather than a pre-canned LUT. I expect to grade 100% of my work, so there is no additional effort for me. However, if you want a fast turnaround, shooting in Neutral with a bit of contrast could be good enough, out of the box. Can't beat that for speed, but it gives you no margin on exposure and WB if you shoot for a final image.

I avoid Cinestyle for low contrast shots (like shooting in the fog) or when the main focus of the shot is medium or light, flatly lit, skin tones. Cinestyle steals bits from the mid-tones and gives them to the extremes. Fewer bits for the skin means more block artifacts, contours, and mottling.

If skin isn't the focus, and the scene isn't low contrast, I like Cinestyle. I used it outdoors with the wrong WB recently, and with the flatter image, I was able to correct it without problem. The extra range means fewer clipped or crushed areas, and that gives me more options and more margin for exposure errors.

BTW, I mentioned diffusion filters in another thread. These generally add light to the whole image and make the blacks milky. I find that I need to crush the blacks back down in post, but the milkiness brings the shadow detail into the range of the sensor, much like Cinestyle does.

[Sareesh Sudhakaran]

Thanks!

Originally Posted by Jon Fairhurst

I avoid Cinestyle for low contrast shots (like shooting in the fog) or when the main focus of the shot is medium or light, flatly lit, skin tones. Cinestyle steals bits from the mid-tones and gives them to the extremes. Fewer bits for the skin means more block artifacts, contours, and mottling.

There's an interesting technique photographers use for low contrast scenes - ETTR (Expose to the right). It is mostly used in RAW mode, but I'm not sure how it performs in 8-bit video.

BTW, I mentioned diffusion filters in another thread. These generally add light to the whole image and make the blacks milky. I find that I need to crush the blacks back down in post, but the milkiness brings the shadow detail into the range of the sensor, much like Cinestyle does.

That's interesting. Obviously it would mean a better result since it was achieved in camera.

I have a question: What if, in one project, you're faced with varying scenarios across scenes - do you use different picture styles? If yes, how do you match them in grading?

[Matt Thomas]

Well Raw is bound to be the best, but I like the Cinestyle over Neutral, but I don't like using the LUT, I'd rather alter the colour myself.

[Giroud Francois]

i do not why you compare raw with other styles in video mode, since you described raw as shot as picture and resized.

[Jon Fairhurst]

Originally Posted by Sareesh Sudhakaran

There's an interesting technique photographers use for low contrast scenes - ETTR (Expose to the right). It is mostly used in RAW mode, but I'm not sure how it performs in 8-bit video.

ETTR is more viable with Cinestyle than with Neutral, because Neutral has the stronger S-shape. For instance, if the face is the brightest object in the scene, you would expose the face brightly. But that puts the face into the shallow part of the S-curve where there are fewer bits and the bits are non-linear. So, for video with people, I prefer to expose for faces using a seven zone system (false colors on a monitor or zebras in Magic Lantern can be used.) If the highlights are too hot, I then adjust the lighting, if possible. I'll push the face a bit darker if there are more highs than lows, but I still try to keep the face centered on the curve.

Cinestyle is more linear, so you can ETTR without losing additional bits in the skin tone. Cinestyle steals bits from the middle, so there's no advantage to centering faces in the histogram.

[regarding diffusion filters] That's interesting. Obviously it would mean a better result since it was achieved in camera.

The picture style is also applied in-camera on the RAW data. Both are viable. The choice of a diffusion filter really depends on the look that you want. For DSLRs, I'd say that you can safely use a Glimmer Glass #1 even when you want a gritty look. The "sheen" of the GG1 is subtle enough that it adds just a touch of magic and tones down the harshest specular highlights. Those highlights alias badly on DSLRs anyway, so you don't lose any "real" resolution.

I have a question: What if, in one project, you're faced with varying scenarios across scenes - do you use different picture styles? If yes, how do you match them in grading?

Yes, I'll use Neutral or Cinestyle as needed. Typically, I'd use Neutral for an interior dramatic shot and Cinestyle for an exterior shot or an interior with high contrast and more equipment (corporate) than people. I don't change styles within a scene. Cutting very different scenes together is always a combination of challenge and forgiving. It can be challenging due to different color temps, exposures, and backgrounds. It is forgiving because it's a different context.

I've got three approaches:
* A global S-curve tool. This is really powerful for giving the highs, lows, and mids just the right brightness.
* Separate R, G & B S-curves. This is hard to use, but I can do it without leaving the NLE. I can make the darks teal and the lights orange, if I want.
* Colorista II. This is easier to use, but I need to open the files in After Effects. This is very powerful. I can easily mask areas and re-light, etc.

In general there are two levels to adjust: a master and a primary. The master is the overall look. The primary is done clip by clip to ensure matching. The third level is "secondary", which is usually applied within masks for re-lighting, or for separating different objects that have similar tones.

[Charles Papert]

While I understand the theory of flat profiles like Cinestyle potentially causing issues in the midrange (as Jon describes it, "stealing bits"), I have yet to see visible proof of this. Anyone have links to a properly done test that can demonstrate the issue in action, as opposed to theory?

Overall, after hearing for the past two years that the H264 codec doesn't allow for much grading before it "falls apart", I've yet to see noticeable artificating in my footage. Throughout a number of DaVinci post house sessions I've asked colorists to keep an eye for this sort of thing as we tweak in looks and it hasn't come up. Granted, I rarely drive things in the direction that a lot of people do, i.e. that blockbuster orange/teal b.s., but I've lifted a lot of stuff out of the toe without issue. Cinestyle has really helped in that regard, to reduce the crush factor in the shadows.

[Giroud Francois]

probably this apply to people working with the native h264 codec.
for sure if you convert first to another, larger codec, like cineform 10 bit 4:4:4 there are chances you can
heavily correct and still see no problem.

[Jon Fairhurst]

I've done casual tests - just shooting my hand on both picture styles under natural, flat lighting. I set the exposure using false colors on a Marshall monitor. I used 320 ISO, as I recall. 5D2. Zeiss. Looking closely, I could see block artifacts and 8-bit contouring on both, due to the flat lighting. The contouring was definitely worse on the Cinestyle stuff. It wasn't terribly worse, but I remember wanting the Natural shot to look better. My reaction was, "cr@p, Cinestyle makes it worse." It's not terribly worse though. But it sure ain't better. I tried grading both and the quantization becomes more noticeable the more you grade it.

It would be interesting to test this with a variety of faces at a variety of ISOs. Theory and experience says that it would be the worst on smooth skin at high ISOs, but I haven't confirmed it over the two picture styles. Codecs have a tough time with smooth surfaces like a balloon, and we know what high ISOs do to images.

In any case, if you want more detail in the shadows and/or more latitude on highlights, Cinestyle is an easy choice. But if I'm shooting a man in the fog in a gray suit, I shoot Neutral.

Still, Cinestyle is my default choice. I only shoot Neutral when skin is the focus, there is no risk of clipping, and there are no important shadow details. I tend not to focus live, and I expect to grade everything, so there are no downsides with Cinestyle for me - except when I want more bits in the mid-tones.

[Chris Barcellos]

Interesting discussion here. I am working with Cinestyle myself, but converting to Cineform initially, and then applying Cineform's Cstyle setting in the Input Encode in FirstLight, developed by Cineform specifically for Cinestyle. The result is what seems like footage that has preserved wide latitude, but not downright flat. This still allows you to apply an additional look in Cinestyle or in your NLE.

Anyone else tried this new feature found in FirstLight.

I am also curious about conturing mentioned by Jon. Jon, do you have a frame example ?

[Jon Fairhurst]

Sorry, Chris, I deleted that test footage. Too busy to set up another test right now...

[Sareesh Sudhakaran]

Originally Posted by Giroud Francois

i do not why you compare raw with other styles in video mode, since you described raw as shot as picture and resized.

Just to check if had more dynamic range. I used the same setting (Cinestyle, WB: Shade) to downconvert from DPP to TIFF. If you find anything lacking in this approach, please let me know.

[Sareesh Sudhakaran]

Thanks for the great reply.

Cinestyle is more linear, so you can ETTR without losing additional bits in the skin tone. Cinestyle steals bits from the middle, so there's no advantage to centering faces in the histogram.

Considering what you mention about using Neutral in low contrast scenes -

If I compare an ETTR shot of a low contrast scene in Cinestyle and Neutral - when I recenter the histogram in post - will the Neutral 'hold' up more in the mid-tones than Cinestyle? Is that what you mean by Cinestyle 'stealing bits from the middle'? Why doesn't this 'mid-tone' issue apply to high contrast scenes?

[Jon Fairhurst]

If I compare an ETTR shot of a low contrast scene in Cinestyle and Neutral - when I recenter the histogram in post - will the Neutral 'hold' up more in the mid-tones than Cinestyle? Is that what you mean by Cinestyle 'stealing bits from the middle'? Why doesn't this 'mid-tone' issue apply to high contrast scenes?

The middle of the Neutral range has the highest bit density. The s-curve makes the bits get thinner at the extremes. Cinestyle has two differences - it's more linear (giving the end regions more bits), and it has lower contrast (spreading the bits a bit thinner across the whole range.) The end result is the the middle of the Neutral range has the highest density of all. It can deliver the most subtlety of the picture styles (Neutral and Cinestyle) and ranges (high, medium, and low) that we're talking about.

If you ETTR on Neutral so that your skin tones are too close to the ends, those face tones could get noisier or more contoured than they need be. Also, if you choose Cinestyle on a low contrast scene, you will likely get more contouring than you would with Neutral/centered.

You asked, "Why doesn't this 'mid-tone' issue apply to high contrast scenes?" Well, it does. But it might be worth the tradeoff to get less clipping of the highs and more details in the shadows.

Charles mentioned that he likes to pull more info out of the toe (blacks). Personally, I'm mostly bothered when the a large area of sky blows out. In either case, if Cinestyle helps us avoid crushing or blowing out large areas of the image, that can be totally worth some loss of subtlety on the mid-tones.

One thing to keep in mind is that nobody (aside from some geeky testers) will play two versions of your film side by side. So relative problems (e.g. style A had more noise than style B) aren't necessarily noticeable, while absolute problems (the entire sky was clipped white) can be a distraction.

With ETTR, Natural, and Cinestyle, there is no one right answer. What's most important is to understand the strengths and weaknesses of the choices, to understand the image that you are trying to capture, and to understand the look that you want to deliver. Put those together and you can get the best result available - even if "best" is still a compromise.

[Charles Papert]

Originally Posted by Jon Fairhurst

Charles mentioned that he likes to pull more info out of the toe (blacks). Personally, I'm mostly bothered when the a large area of sky blows out. In either case, if Cinestyle helps us avoid crushing or blowing out large areas of the image, that can be totally worth some loss of subtlety on the mid-tones.

I'm pretty used to working around hot skies, that has always been an issue with video. What is disconcerting with these cameras is that the specific advantage of extended shadow detail that emerged with digital and later HD acquisition became reversed, as all of the picture styles native to the camera tend to crush. While that works fine for a lot of indie folk who favor the contrasty, Magic Bullet Looks style (strong s-curve), it's not a look that I strive for unless specifically dictated.

Certainly I'm not a fan of blown skies but while that can be generally managed by exposure, fill, filtration or other means, I have a tougher time managing the Ektachrome-like toe present in the native profiles so the Cinestyle has been a welcome relief. However--the vast majority of the footage I have shot with these cameras has been, fortunately, in relatively controlled environments so I haven't had to come up against the dreaded sun/shade high contrast situations all that much.

Regarding diffusion filters and previous statement notwithstanding, I have had a couple of daylight situations where I have been forced to shoot in shade, under dappled sunlight which of course is a contrast nightmare. On a pilot for FX, I had an all-day shoot in the forest where I didn't have the luxury of moving the actors into the best light as required, due to a built setpiece with specific orientation. I had a good amount of grip and electric support to get me part of the way there but there's only so much you can build up exposure in the shade before it looks too fake. For quite a bit of the day I was able to block out the direct light via 20x20's but for the middle part of the day, I just had to go with it. I decided to go with strong black promists the whole day as that would "take the curse off" the highlights a bit (by spreading the highlight out a little to avoid the clipped look). Since this was a spoof science fiction show and already pretty campy, I felt that the resulting diffused 70's/80's look would work fine in context, and in fact the creators loved it. It's not a technique I'd use everyday, for sure. Incidentally, the first half of that clip (shot in an actual submarine, god help us) pretty neatly demonstrates the black crush issue--this was pre-Cinestyle, I shot on the Neutral setting with contrast and detail all the way down. I feel pretty strongly that Cinestyle would have given me a lot more control in both the interiors and exteriors for this project.

BTW Jon--you said a few posts back that you tend not to focus live. Presumably you aren't shooting narrative...?!