Film "look"

Serena wrote on 9/6/2005, 11:11 PM

One of our very popular discussion topics is "how to achieve the look of film". Generally the discussion very quickly centres on 24fps, strobing of movement, film grain, depth of field (and even scratches). Certainly some of these effects will cause people to deduce it must have been shot on film. Of course, if you're going for cinema release via film, then you have to get to 24fps.

Having come from a film background I'm forever saying "it's none of those things, it's all about the response characteristics of photographic emulsions compared to CCDs". The most obvious message that "this is video" are burnt out highlights (clouds, breaking waves, etc). On going to video they were the first things that annoyed me. Expose to keep highlights within range and the lowlights get all clogged. So, graded filters and all that stuff. I'm sure I've never convinced anybody here except those already of the same view.

On the CML site mention has been made of a paper that nicely discusses this point: Michael Bergeron; "Increasing Dynamic Range for Digital Cinematography". It opens with Kodak's published experimental results showing that high definition video delivers a dynamic highlight range relative to an 18% grey of no more than 3 or 4 stops, compared to 15.9 stops for some colour negative films. This paper provides a good grounding for any discussion of "film look"; I think it's a worthwhile read. See:

http://www.hpaonline.com/i4a/pages/index.cfm?pageid=236

Comments

Jonathan Neal wrote on 9/8/2005, 1:09 AM

Thank you very much - you have really summerized the overall thoughts of those looking for the "film look". Now, only to find the "for dumbies" version =) Thanks though, this is useful.

Spot|DSE wrote on 9/8/2005, 7:06 AM

Very useful, Serena,thanks for pointing to it.

Serena wrote on 9/9/2005, 12:10 AM

The general parts of the paper are the most useful to us. Once Bergeron gets into the specifics of the FILM REC mode of the Panasonic Varicam then he's talking about things over which our cameras give us little control. The "cinema-tone" option in the Sony cameras provides a "film like" response curve along the lines discussed in the paper, but I've not seen anything that specifies what characteristic it actually gives. The paper makes us aware of the issues pertinent to video viewing standards (compared to film) and of course Bergeron is talking to DPs shooting HD for transfer to film.

Obviously if we expose to keep all scene brightness elements within the range of our CCDs (or CMOSs) then in post we can manipulate things to give a look we like. This means being careful with exposure and in most scenes using filters (eg, graded) and other control elements (eg, reflectors). In the final product the film look is about the perceived tonal range under relevant viewing conditions, without clipped whites and blacks.

Grazie wrote on 9/9/2005, 12:47 AM

Thanks Serena.

Your - "Obviously if we expose to keep all scene brightness elements within the range of our CCDs (or CMOSs) then in post we can manipulate things to give a look we like. This means being careful with exposure and in most scenes using filters (eg, graded) and other control elements (eg, reflectors)." . . has vindicated my recent & early approach and adoption in using grad NDs and Polas.

I often felt I needed to video a scene as good as I could. This meant having an understanding and realisation what I would like when I previewed my taped footage. NOW, using filters, and bouncing light with reflectors, I can have a better initial "stab" at what I might want to do with a "look" in post. Previously, if I had captured something approaching what I wanted, then in post I could get nearer but not close enough - plus the accompanying graining, plus the the effect of smoothing.

I now use a small arsenal of NDs - 0.3 thru to 0.9 grads and solids; linear polarizer; supermist black; warm tone enhancer and a tabac filter - because I wanted it! I'm attempting the brightening the darks and bringing down the brights but holding the colours. I'm wanting a "hot" natural look. I often see people on video visually looking "cold" in shirtsleeves! And it is obviously a gloriously hot day.

Thanks again for the post,

Time to read in full now!

Grazie

Serena wrote on 9/12/2005, 12:02 AM

This addition to the topic is rather "OFF TOPIC" for Vegas, but seems an appropriate place to put this -- I hope that it is of some interest and that you'll forgive me!

Just playing around with the HDR-FX1E cinetone preference and various tests indicate that it does indeed handle highlights in a much "friendlier" manner. Obviously DSE will be quantifying all this stuff in his forthcoming book and has the facilities for doing it properly, but nevertheless here are some numbers that will show you the sort of characteristics. I've compared PP1 ("appropriate setting to record in the HDV format") and PP4("appropriate setting to record film-like pictures"). Both set to cineframe. Instead of measuring video-out voltage (which would scare me to try) I recorded a constant white screen at various f/stops and used the Vegas waveform monitor as indicating video out level. Pretty crude, I know, but the results were encouragingly consistent.
So rather than posting a graph I'll let you plot your own with the following numbers. If anyone wants a graph I can email it to you. Light level means light intensity on the CCD and is relative to that at full aperture.

light 100 ; 64 ; 32.6 ; 16 ; 8.16; 4 ; 2.12
PP4 105.5 ; 101.5; 87.5; 59 ; 35.5 ; 20 ; 12
PP1 104 ; 101 ; 96.5 ; 69.5 ; 45.5 ; 27.5 ;16

Recording was done with aperture changed in opening and closing directions and was very consistent. Zebra was set to 100+ and was way over for the first two points of PP1. OdB gain, 1/50 sec, no ND, all manual. Screen was illuminated to start seeing small zebra pattern at f/2 for PP4. Very nasty clipping in PP1 when over exposed.

I repeated with neutral density filter in but without increasing illumination of the screen. I'm going to repeat this at greater screen brightness because my results indicate that putting in a ND changes the processing characteristic (not just a constant cut of 2 stops (say)). This surprised me and I want to confirm this and see what happens in the highlight region.

craftech wrote on 9/12/2005, 6:49 AM

Serena,
How are you handling artificial lighting (particularly stage lighting which varies constantly) as opposed to natural. It would seem to compund the difficulties of your approach?
John

GlennChan wrote on 9/12/2005, 3:16 PM

Serena, how about displaying a gradient on your television?

In Vegas, be sure it's from 16 16 16 RGB to 235 235 235 RGB to avoid illegal colors. Use pan/crop to zoom out on it so stuff doesn't get cut off in the overscan areas.
Put a grey (i.e. 128 128 128 RGB) underneath.
You will get a little experimental error from TV (mis)calibration and ambient light hitting the television, but that's acceptable.

Serena wrote on 9/12/2005, 8:43 PM

Hi John,
You've identified a major practical difficulty -- often we don't have enough control over the situation and almost never if shooting "on the run".

What I'm attempting is "understand the tools". I come from film and am new to video and I get pretty frustrated by the paucity of technical data provided with prosumer products. Understanding characteristics and limitations is a preliminary for optimising results. The Sony HDV camera, for example, provides "picture profiles" that the user can engage and each has parameters that are user-settable. What aren't defined and which cannot be modified are the camera's signal processing algorithms. You can go out and video stuff for visual comparion of various modes, but this is rather qualitative and I find not sufficient. In fact I did this when I got the camera and dismissed "cinema-tone" as a waste of time (looking for the wrong thing).
Constantly people are seeking the "film look" (not everyone, I admit) and unless we understand what that is then we can go off on wilder pursuits than Jason seeking the Golden Fleece. In general the problem is that our video sensors have an inadequate dynamic range (8bit instead of the 12 to 16 needed). However where people don't see this as a problem, then there isn't one (for them).
As an aside, in any discussions of technically related issues great confusions arise and many blind alleys are followed when contributions don't differentiate between "fact", "opinion" and "guess". I'm trying to get a handle on some facts; I'm most happy to be instructed by people who know, questioned by anyone, and engaged by opinions. Certainly none of this is definative.
Serena