The Film Look on DSLRs

[Jon Fairhurst]

Originally Posted by Sareesh Sudhakaran

Stills or video - it is the same dynamic range from the DSLR sensor under controlled lighting conditions. The zacuto videos pointed this out clearly.

Funny. I got the impression from the Zacuto tests that the DSLRs clip much sooner than film - especially from the 2010 test on the light bulb. Fortunately, the Canon DSLRs clip cleanly at white, rather than with a yellow tinge. I shot some video recently with Cinestyle that had practical lights in the scene and it still clipped with a bit of a hard contour line. Film would have clipped more gracefully.

Let me ask my question in another way: Was there a situation in which you tried your best to emulate the film look (in your own subjective opinion) and failed, even though you satisfied all the established conditions (or norms) to attain that look?

Well, there's the above situation with practical lights in the scene. It was a subtle fail. Selective filtering of the hard edge could probably make it less visible.

And, frankly, DSLRs look very nice to me, but never quite look exactly like film. It's a subtle difference. I think the problems have to do with limited horizontal resolution, vertical aliasing, 8-bit depth, and compression. Eliminate those problems - say by shooting a RAW timelapse - and add just the right film grain, and I think the result can be indistinguishable from film.

Well done noise reduction can help with the video compression and 8-bit artifacts, but the result can get a bit plasticky. It's not a bad look, but I can clearly see that it's not film.

The film standard is a challenging one to ask about. I can't do double blind tests on my own material. And the film look is subjective. I think you need to do your own tests and decide what looks enough like film to you.

[Sareesh Sudhakaran]

Originally Posted by Jon Fairhurst

And, frankly, DSLRs look very nice to me, but never quite look exactly like film. It's a subtle difference.

The film standard is a challenging one to ask about. I can't do double blind tests on my own material. And the film look is subjective. I think you need to do your own tests and decide what looks enough like film to you.

Thank you, Jon, for the comprehensive answer. I, too, feel quite a difference when shooting stills and video with the same settings (neutral mode on the Canon).

The only reason I raised this thread was because of the nagging doubt of whether, assuming the histogram never clips, video will look like film or not. Guess I'll have to find out for myself! Thanks again.

[Sareesh Sudhakaran]

Originally Posted by Chris Barcellos

Here are films I have tried these different techniques over the years, most of them for DVinfo contests. The latest was shot with Technicolor Cinestyle. Point is you may think they are not filmic, while I may think I have accomplished my goal.
]

Thank you for sharing your videos. I enjoyed Perfect planning. There's one point at around 2.20 when the agent's face first appears, and we see all her wrinkles and highlights. Do you think, with creative makeup and
lighting with nets (or whatever) the effect would have been better?

I have Cinestyle, but my aim here is to get the look while shooting. I did try cineform, and found it useful to a certain extent. However, in my opinion, transcoding all the footage of a feature film is not the best way to go about it in the digital age. But that's another matter altogether.

Actually, I've seen a movie shot on 16mm that looked like video - on DVD of course. It's possible they used the worst lenses, then telecined it on the worst system with the wrong settings.

[David St. Juskow]

Sareesh, I'm a little confused by what you keep referring as "film look." First of all, nothing looks like film other than film. The silver grains that react to light chemically have no equal in the digital world. One can argue aesthetics all day, but there is simply nothing digitally that matches the random shapes and patterns created by silver halides reacting to light. You're comparing random shapes to a grid of pixels.

Added to this, is the amount of frames per second that the eye is detecting to produce the illusion of motion. With film, that's 24 frames- a very crucial part of the "film look". There's also the dynamic range of film stocks, which, to my eye, still beat out digital video, even beloved cameras like the RED. If you watch a projected 35mm film vs an HD one, there's no doubt as to which has the richer blacks, and the wider amount of detail in both extremes (lights and darks.)

So the best you can do to achieve this look with a DSLR is to light it well, shoot at 24 fps, and play with the image in post to approach the way film reacts to light. You can also do this in the camera settings somewhat, and there's lots of debate about the best camera settings to use- something I haven't really experimented with, so I can't be much help there.

[Edward Mendoza]

This was shot on a 60D and treated in post. It's not film, but with a little tender lovin' care, you can get fairly close.

Grandpa and the Kids - Canon 60d Slow Motion on Vimeo

[Sareesh Sudhakaran]

Originally Posted by David St. Juskow

Sareesh, I'm a little confused by what you keep referring as "film look." First of all, nothing looks like film other than film.

But we still have to emulate the film look while making feature films, unfortunately. Hence, the discussion.

[Sareesh Sudhakaran]

Thank you for sharing, Edward.

[Sareesh Sudhakaran]

Here's a great example of the film look with Leica-R lenses on a 5D Mark II:

"Courage" Official Trailer on Vimeo

[Jon Fairhurst]

To me, it's the grading that gives it a film look. The colors are somewhat muted with the mids pushed toward orange/red. It's also due to the art direction - there are no blues aside from sky and water, and they are especially muted - but not to the point of looking odd. The highlights and blacks are pushed, but to a limited degree - there is a lot of range remaining in the mid-tones. My only beef with the color correction is that the skin tones could have been conformed a bit more consistently. I would have muted the greens to give it a drier look, but that's an artistic choice.

Aside from that, it's just a matter of using good production values and glass. But I don't think that you can get that look right out of the camera.

Note that certain items were saturated - the old guy's scarf and the woman's purple dress. To get those to pop consistently requires a good grade.

Also, note that this is a nice look for a western. I wouldn't be right for, say, sci-fi, unless they were on a desert planet. Or if it's Cowboys vs. Aliens. ;)

[Gabe Strong]

Originally Posted by Sareesh Sudhakaran

Stills or video - it is the same dynamic range from the DSLR sensor under controlled lighting conditions.

Um....I don't think this is true. DSLR's get better dynamic range in 'stills' mode than they do when shooting video. Maybe I'm wrong, but
that's what it appears to me, just looking at it with my eyes.

Now I can't verify that what they are saying is true, as I haven't measured it with a scope. I just know what I have seen, and it appears to my eyes that
there is a difference in dynamic range between what you get out of a DSLR in video mode and still mode.

[Jon Fairhurst]

According to the guy who did the test, he used "the flattest contrast settings possible." But the test was done in 2010, so I would guess that he used the Neutral or Faithful picture style. It's possible that one can get more DR using CineStyle.

ProVideo Coalition.com: Stunning Good Looks by Art Adams

There's DR and then there's DR. While the camera might get 10 to 12 stops of light (real world DR), it only has eight bits of resolution on the digital side. When you spread those bits too thin, banding is the result - especially after grading. Film doesn't have banding. ;)

[Roger Keay]

Typical DSLRs have 8 bits of depth in the video or JPEG images versus 14 bits in a RAW file. The camera captures the dynamic range but has to process the image to 'fit' into the 8 bits of video output. The usual approach is non-linear handling of the darkest and lightest parts of the image in much the same way as film has a non-linear response curve.

If you process still images from RAW files you can work with the full bit depth and use curves to put the light and dark values where you want them. The remaining problem is actually displaying more than 8 bits of dynamic range - most displays can only handle 8 bits. Some professional monitors and video projectors can handle more than 8 bits of digital data and accurately display the full dynamic range.

If you want more dynamic range you need to use a camera that delivers more bit depth - typically in the form of RAW frames (unprocessed sensor data) that can be manipulated heavily in post production without showing excessive noise under difficult lighting conditions.

[Sareesh Sudhakaran]

I have a new definition for the 'film look'. It is - anything that doesn't look like video. If a layperson can tell - it fails the test.

Originally Posted by Jon Fairhurst

there are no blues aside from sky and water, and they are especially muted - but not to the point of looking odd.

In your experience, is blue the primary culprit on why video looks like video? I always observed how the whites always looked 'video-ish'. I had a theory during the shooting of my first film - avoid whites at all costs! Maybe it was just the blue channel?

Aside from that, it's just a matter of using good production values and glass. But I don't think that you can get that look right out of the camera.

Agree with you 100%. The author of that video talked about the 'creamy' look he got from those Leica lenses. Now all I can think of is creamy. Is that really a property only of the lens? If so, can it be matched by top glass from Canon or Nikon?

Note that certain items were saturated - the old guy's scarf and the woman's purple dress. To get those to pop consistently requires a good grade.

Yes, no doubt about that. There were good people behind the camera there.

Also, note that this is a nice look for a western. I wouldn't be right for, say, sci-fi, unless they were on a desert planet. Or if it's Cowboys vs. Aliens. ;)

True. Coming back to the 'blue' question: is that why it wouldn't work on a 'colder' film?

[Jon Fairhurst]

Regarding the "blue question", yeah. For an antarctic film, you'd really want to push the blues while just barely keeping the skin tones from looking unnatural. But you'd want the people to look cold to the bone. The exception is the lips. Often in blue/green films, the lips and blood are still very red. In this case, I'd want to do some exhaustive tests with makeup. You'd want to make it so that when you color correct for a snowy landscape that the faces and lips are close to where you want them. And you want the lips redder than the face, so you can adjust them without rotoscoping a mask.

On the flip side, when you go inside of the "Ice Station", I still want the walls to be cold and blue, but I want the faces to be warm and red/orange.

FWIW, here is a short that we did in mid-December, 2008 - about a month after the 5D2 came out. But first, a bit of history...

* The original 5D2 firmware was auto only. We had to untwist the Canon lens, shine/shade a light into the camera and hit exposure lock at just the right time to attempt to set the exposure.
* Quicktime didn't understand Canon's 0-255 range and would clip at 16-235. Blacks weren't just crushed, they were chopped!

For this shoot, we had a Nikon 50/1.8, a Canon 70-300/4-5.6, and no ND filters. We programmed a custom Picture Style that attempted to lift the blacks from 0 to 16, and that reduced our bit depth. We used an ExpoDisc to set the white balance, but the WB changed minute to minute as the clouds changed and as we moved the direction of the lens. Shooting in the snow is tough! We color corrected in 8-bits, so the banding is horrible. Such were the early days of budget shooting with a DSLR! (And, no, it doesn't look like film, but it's a cold grade outside and a warm grade inside.)

The Last Outpost on Vimeo

Regarding lenses, nice detail and a creamy feel is the holy grail. :)

That 70-300/4-5.6 was anything but creamy. Photos of wildlife showed the fur as "crunchy", especially when cropped. I sold it and got an EF 200/2.8 L II. What an improvement! I think it comes down to micro-contrast. The 70-300 didn't co-locate the transitions from different wavelengths, making edges look ratty and distorted. It had a plastic feel. By contrast, the 200 produces super-clean edges. I see the same clarity from Zeiss glass. I liken it to the difference between an audiophile's tube amp and a budget transistor job.

There's an additional trick for a creamy look, and that's to use a diffuser. I have Glimmerglass #1 and #3 filters. Unfortunately, the #1 is too subtle and the #3 too aggressive for the perfect film amount. The right choice depends on the content. I didn't use a diffuser at all for a recent technology piece. I used a #1 on an interview with male talent. (And I would have used a #2, if I had one, but they are aren't available in round formats.)

There are a few benefits from using a diffuser. If it's subtle, it adds a Hollywood glow to the image. It knocks down small imperfections and makes skin look better. It can reduce aliasing to some degree. It won't work miracles though. By the time all the aliasing is gone, the diffusion will be over the top. A mister can also help in delivering that Close Encounters, streams of light, look.

FWIW, I'm pretty sure that the Western example you showed didn't use any diffusion. The white stubble of the beard and reflections on the skin detail on the foreheads was really sharp - and over-sharpened for my taste. A subtle diffuser would have tamed that a bit. But it's an artistic choice. Keep diffusion low for grit, though a GG#1 would still work. As I say, the #1 is VERY subtle. I'd like a #2 as a standard film filter. A #3 starts to become apparent as an effect. There are also #4 and #5 filters for when you want to produce a true effects shot.

Of course, you should only use a high quality diffuser. Why spend the money on a lens with great coatings only to put a bad piece of plastic in front of it? You want to retain the micro-contrast between wavelengths while blurring the hot spots slightly.

[Sareesh Sudhakaran]

Originally Posted by Gabe Strong

Um....I don't think this is true. DSLR's get better dynamic range in 'stills' mode than they do when shooting video. Maybe I'm wrong...

No, you're not wrong. RAW from the camera has far more dynamic range than the video. I tested it today (crudely), and I discovered the difference might be as great as 3-4 stops...Weird.

Here's a link to my results: Dynamic Range comparison of RAW v/s video mode on the Canon 550D | Sareesh Sudhakaran