Most people agree ‘the film look’ is the holy grail. In my comparison of cameras based on dynamic range (which many believe is the prime cause of why video looks like video) we have seen how the Arri Alexa is the camera that comes closest to achieving this dream.
But what about all the cameras that are under $20,000? What about DSLRs, the BMCC, the Sonys and the Canons? You’ll find infinite plug-ins and tutorials on how to get the elusive ‘film look’ but do they deliver? Can they deliver?
Defining The Film Look
This is where most people get stuck. I’m no different. To make it easy I cop out and say: As long as a lay person cannot tell the difference, it’s fine. Most of the time though, after I’ve deluded myself sufficiently that my current project looks like film, I show it to a few people and ask: “Was this shot on film or video?”
It’s not a problem that digital video has, inherently. Magazine stills are shot on digital cameras and at printed in 300 dpi. Most of the world did not know or realize when they made the transition from film to digital. A cheap Canon T4i is quite capable of shooting mind-blowing stills in 14-bit RAW – so there’s nothing wrong with the camera or the sensor.
Is the Canon T4i good enough for theatrical distribution?
When most filmmakers start out to buy their first video camera they are limited by budget. The question one is most likely to hear is “Is so-and-so camera good enough for theatrical distribution?”
I asked the same question when I bought my first video camera, and I did everything I could to convince myself it did. But it didn’t, did it? This was in 2008. Have things changed today?
First, let’s look at the numbers:
|DSLRs/Prosumer||DCI 2K Flat Distribution|
|Color Bit Depth||8||12|
|Color Space||Rec. 709||CIE XYZ/DCI P3|
|Chroma Sub-sampling||4:2:0 (4:2:2)||RGB|
|Data Rate||24 to 100 Mbps||250 Mbps|
Resolution-wise, most video cameras are good enough for theatrical distribution. Even the DSLRs with line-skipping achieve about 700 vertical lines of resolution, which is equal to 16mm film. Since 16mm film can be blown up for theatrical distribution this is a non-issue.
Codecs in both categories are consumer-grade display codecs, but what separates the two is the data rate. You will notice that DCI has a higher data rate than broadcast television, but is it good enough? What’s the true story here? Let’s do a quick test:
From the article on the costs of working with uncompressed video we can figure out that an 8-bit 1080p image (according to the first column) is being compressed at about 7:1 (for intraframe) and about 20:1 (for interframe). The DCI codec (intraframe) is compressed at the rate of 7:1.
The DSLRs and Prosumer cameras, shooting in 8-bit has the same compression as 12-bit DCI! To know why both interframe and intraframe codecs equal each other, read my article on Interframe vs Intraframe compression. Data rates don’t tell the whole story, do they?
What about compression? Surely H.264/AVCHD is the culprit? I don’t have a clear answer, but ask yourself this: If a DSLR sensor can shoot great stills, and the data from this sensor is being converted to 8-bit H.264 in camera – how is this any different from film scans transcoded to 8-bit H.264? Surely, transcoding itself doesn’t introduce the ‘video look’, does it?
You see, there is nothing technically incorrect with the specifications of cameras, so you can rest assured your camera is capable of meeting the specifications of theatrical distribution. Broadcast, is another matter entirely, as I’ve written here.
This is why I decided to limit my definition to just the ‘final result’: Can a lay person tell if your movie was shot on video? If yes, you have not achieved the film look.
How does one know the difference between film and video? Film has blown-out highlights, too. Film has low-contrast images, too. Film can be desaturated in color, too. Film can have low resolution, too.
What people do to get the film look
In this section let’s take a look at some of the ‘common sense’ advice to get video to look like film:
Most modern cameras in every budget are capable of 24p, so this is no longer a factor. However, some camera sensors first encode in 50i or 60i and then convert data into 24p. Interlacing is a dead giveaway.
DSLRs have electronic shutters for video, which can be set to replicate the effect of filmic motion blur. The most expensive cameras like the Alexa and the F65 have mechanical shutters, but what difference do they make? Here’s a lesson from Red that provides a simple overview of the difference.
Depth of Field and Bokeh
Film grain isn’t rectangular, whereas anything in video must be rectangular, and that includes video noise as well. There are products in the market today that try to simulate film grain. On an aesthetic level, I can imagine why one would want to add grain, when Kodak was trying its best to get rid of it. However, video with grain still looks like video, no matter what.
Cameras like the Canon T4i shoot great stills even in JPEG mode in 8-bit, like this one shot on JPEG, color corrected, cropped and recompressed for the web:
As we have seen in Color Space and CIE XYZ, sRGB and Rec. 709 are similar for all practical work. If 8-bit sRGB still images are okay, what’s wrong with 8-bit Rec. 709 – they are practically the same thing!
Similarly, movies shot on film, scanned to digital and then compressed to Rec. 709 for consumption don’t look like video. If the original film had colors beyond the gamut of consumer displays, their loss should be clearly felt. Are they? Did you catch a loss in colors the last time you watched a Blu-ray movie?
Clearly, adding or changing color is not the key to achieving the film look, yet you have plugins and tools that do just this.
This is one aspect even camera manufacturers take seriously, by trying to ‘flatten’ out the image via Log Gamma or whatever proprietary name they want to call it. On DSLRs, we have Technicolor Cinestyle. My tests with Cinestyle over Neutral (the built in setting) show that there’s no worthwhile difference.
Even on the C300, I never use Log because I can get what I want with Rec. 709! Here’s a test of Canon Log with Canon LUT applied:
Even if you haven’t played the above video, can you tell whether it was shot on film or video, by just looking at the still frame? However, we have all probably seen Mobius:
Mobius is the closest to the film look I’ve seen, yet there are quite a few shots in which you can clearly tell it was shot on video. The brilliant grading and camera work cannot hide this.
So, is my camera really good enough for theatrical distribution?
Well, I already told you that it is, technically.
Yeah, but can I apply all the tools at my disposal to make it look as good as film?
You can come a lot close nowadays, but you can’t beat film.
Wait. Why don’t you ask me what you really want to know?
Okay, will distributors, the film festival screening committee, the film critics or the moviegoing audience know the difference?
Yes, they will.
Will they care?
I would love to say they won’t. I would love to say content is king, and if you make a great movie nobody will care. But in reality I have learnt the opposite – people do care about the look. Especially the people who guard your entry to theatrical distribution. And the movie going public, too, for that matter. They might not be able to articulate it, but they will know.
Don’t take my word, ask your own friends and family if your masterpiece was shot on film or digital. You will have your answer. And before you go investing in some camera because somebody online promised you the film look, look at footage from it and show that to your friends and family.
But why is that some footage looks like film while others from the same camera don’t? Why aren’t the results consistent?
Because you can hide many of the flaws, but not all of them. If you’re shooting an interview or well controlled closeups, you can actually deliver stellar results, as Mobius shows. However, a movie is more than a bunch of closeups. There are many shots that have to be intercut with each other.
Isn’t this the bane of the low budget indie filmmaker anyway? You hardly have enough time on a shoot to get all your shots, let alone get them Hollywood-style.
Do you have 60 to 90 days for a shoot, or is it more like 15 to 20? Can you control the lighting and dynamic range of every single shot? Do you have the post production resources to grade your movie under the care of a genius colorist?
Anybody can download Resolve nowadays but how many can use it to its full potential? Out of this, how many have the ‘artistic eye’? If you really have the budget to grade in such an environment, why not use that money to shoot with a better camera in the first place?
But let’s not lose hope and get carried away. The film look isn’t everything, nor should it be. The bottom line is one can still make great images even with a cheap DSLR like the Canon T4i. You should be proud of these images.
But don’t ever kid yourself, or let anyone else sucker you into thinking it will look like film. Ask me again, now.
Is my camera good enough for theatrical distribution?
Yes. Here’s a movie shot on DV that did great:
And here’s one from the same director, with a 2K camera without cinematic DOF:
Okay, wise guy. You’ve told me my dream camera can never look like film, so what causes the video look?
I don’t know. But I have a theory.
Do cameras technically qualify for theatrical distribution? As we have seen, there’s nothing to prevent it.
Do cameras sub $20,000 offer the elusive film look? The answer was in the negative.
Let’s look at each camera segment to see at what point we get a consistent sequence of images that can pass as ‘the film look’.
Let’s start with DSLRs.
The following is a table of eight videos, one for each well-known DSLR. The names might be in the videos themselves, but the goal is to watch them without prejudice, and decide for yourself whether every shot is filmic or video-ish.
Important: This is no reflection on the content or the effort these filmmakers have put in.
Of course, it goes without saying that there might be better or worse examples out there, but the point I’m trying to make is that when shooting in extremely low budgets, don’t depend on your camera to cover up for your mistakes in lighting, production design, etc.
The last video is from the BMCC, a camera that looks very similar to what we see on higher end cameras. It’s for good reason I gave this gem of a camera the title of Best Video Camera for Web Videos. And, don’t forget the Master Guide to Rigging the BMCC, in which I cover some of the reasons filmmakers with lofty ambitions shy away from its sensor size.
Cameras from $5,000 to $10,000
Here are four shorts, from cameras between $5,000 and $10,000:
Again, it goes without saying this is no reflection on the content or the effort these filmmakers have put in. Can you spot the ones shot on video? Is it that obvious? How well was it hidden in some shots?
I’ve already included one camera that costs $10,000+ in the DSLR category, the Canon 1DC. But it is just to highlight its differences, if any. It’s only when you compare images side by side that some aspects become clear.
My pick? I’m not a fan of any of the results from these cameras, but I prefer the Sony FS100 over the others.
Cameras from $10,000 to $20,000
So, which ones look like film and which ones look like video, even it is for a couple of shots?
I have to say I find nothing in these cameras special enough to warrant their price. I shoot with the Canon C300, because it offers me the fastest workflow, with decent image quality. Then again, I’m not shooting fiction with it. I don’t think I can, unless someone pointed a camera at me.
Cameras over $20,000
Here we compare the big boys: Red Scarlet and Epic, Sony F55 and F65, Arri Alexa and the Canon C500:
See anything you like? Except for the Canon C500 and the Sony F55, which I’ve compared here, the rest are already noted cameras capable of pleasing mass audiences. At least, I feel they are good enough to match film. We have seen in the dynamic range comparison of cameras that most of these cameras are close to or above Kodak Vision3.
Of all the cameras compared here, the one I find truly delivering ‘the film look’ in-camera, without the need for digital voodoo, is the Arri Alexa.
So, what’s my theory on getting the film look?
It’s just a theory. Many say that shooting RAW is what gives you the film look, and that is correct to an extent. But it is possible to get the film look without RAW.
You know the famous Sherlock Holmes’ saying: Blah blah blah…whatever remains must be the truth. Well, what’s left? If you look at all the cameras that do pass muster, and the ones that don’t – one difference stands out quite clearly:
The cameras that deliver the film look don’t use chroma sub-sampling for their film-like imagery.
Like this one for example:
This was shot on the Sony HDC-F950 in 4:4:4, without RAW. As you can see, RAW isn’t required. We saw this earlier when we discussed in-camera JPEGs vs RAW stills.
Maybe soon one day, I’ll wake up to the news: Chroma sub-sampling banned! Then it might be possible for any camera to deliver the film look, and this article can lay to rest.
To know more about the science behind film vs digital, and why the film look eludes most video cameras, check out this brilliant white paper from the BBC: The Film Look: It’s Not Just Jerky Motion… by A.Roberts.
Now you know. Do you think it is chroma sub-sampling that is causing video to look like video? Tell me.