David Kong here, with my second post in this video/article series that is all about working with less gear, for less money, and making it look good.
This video goes very in-depth with the DSLR – I cover everything you need to know about formats, codecs, and dynamic range in order to get the most cinematic images. I’ll be explaining how I creatively worked around the limitations of this low-end DSLR with a pretty bad codec and pretty bad dynamic range. Most of my techniques in this video are about preserving as much of the image as possible, to allow for the most flexibility when color grading.
I shot this particular film on the Canon T2i/550D, so the discussion is framed in those terms, but all of the principles apply equally well to most DSLRs and many other video cameras.
Part Two: All about the Camera
Part One: The Gear
The first video/article was all about the gear (lenses, filters, tripods, etc). If you haven’t seen it yet, check it out here:
The First Article: Working with Less (or, Get Out There and Shoot!) Part 1
The Proof of Concept:
I’m using one of my films as an illustration for all of these concepts, so if you haven’t seen the Portrait of Macerata, check it out here. I’ll be talking about that film all through this series.
Below is an overview of everything I cover in the second video
Viewer question: How did I approach the shoot? How much did I plan?
Viewer question: Did I get signed release forms from everyone in the film?
Viewer question: How did I work around moire?
CineStyle is by far the most important tool in my fight against the video look. Clipped highlights and crushed blacks are the biggest giveaways that you’re shooting in video and not film or high-end digital cinema. The “video look” is very contrasty, and usually with clipped highlights. CineStyle is a color profile that manipulates the data coming off of the sensor BEFORE it gets compressed with the codec, and it helps you avoid compressing your highlights and shadows.
CineStyle is one of several presets that you can stick on your camera to get rid of that strong contrast curve and preserve as much data as possible, and most of them work very well. A lot of it, really, is a matter of taste. I prefer Cinestyle because I know that the people behind it (Technicolor) really know their stuff. Another one to know about is Stu Maschwitz’s Prolost Flat. He just works off of the “neutral” preset that comes built-in to the Canon cameras and tries to get the flattest profile with that. I’d recommend trying them both.
Codecs are super complicated (and super interesting!) so I’ll just boil the explanation down to several main points about codecs and explain why you need to understand these concepts.
What codecs do
A codec is like a funnel. You’ve got all of this data (raw coming off the sensor), and you have to fit it into a smaller container – your memory card. So the codec is what shrinks the size of the data stream coming off your sensor down to a size that can fit on your memory card. The main reason why you have codecs is to save space. Raw video data is HUGE, and memory card transfer speeds can rarely cope.
Most video codecs use lossy compression, which throws away some of your data to make it fit onto the SD card. There are dozens of different codecs and dozens of different techniques for how they can decide what to throw away and how to do it, so that it has the smallest impact on the image as possible.
This concept doesn’t directly affect that way that I shoot, but it’s a very important part of how codecs work so it’s relevant for me to talk about real quick. You’ve probably heard the phrases 4:4:4, 4:2:2, and 4:2:0. That’s chroma subsampling.
When we’re storing pixels in a codec, the brightness value and the color value are separate. Brightness is called luminance, and color is called chrominance. Researchers discovered that our eyes are much, much more sensitive to changes in brightness than to changes in color. So they figured out that, if you keep all of the brightness information but throw away some of the color information, your eye doesn’t really notice.
- 4:4:4 keeps all of the information.
- 4:2:2 throws away 1/2 of the color information (keeping all of the brightness).
- 4:2:0 throws away 3/4 of the color information (keeping all of the brightness).
Canon DSLRs use 4:2:0, the worst kind. But we survive It’s really not an issue unless you’re planning to do FX work like greenscreen. Trying to pull a good key from a 4:2:0 codec is a nightmare!
Bit Depth is the range of values that you have to work with – the number of possible colors. The higher bit depth you have, the more in-between shades you have. Low bit depth can be a problem when you have a very long, smooth gradient. You’ll see a weird “banding” pattern sometimes because there aren’t enough in-between values in the gradient.
These cameras use 8-bit. Not great, but we can get by.
The codec we’re using: h.264
Almost all DSLRs except for the high-end ones use a codec called h.264, which is designed to be a delivery codec. It throws away as much as it possibly can while keeping the current image the same. But what if you want to change that image? What if you want to make it darker, or lighter, or change the color? Well now you want some of that information that got thrown away, and you’re stuck. The huge advantage to h.264 is that the file sizes are very small, but the disadvantage is that the codec is NOT designed to be color corrected.
h.264 is doing a whole bunch of different compression tricks at the same time, but there are two that are particularly relevant to us because the way that you shoot can have an impact on the way that the camera compresses.
h.264 has two principles you need to know about:
- Encode for the bright things.
- Encode for contrasty things.
The camera will encode bright and contrasty things fairly well, but it will give a lot less attention to areas that are not bright or contrasty. That’s great for the image as you shot it (because our eyes are more attracted to bright and contrasty things), but what if you want to change the exposure? What if you want to brighten the image, for example? Now we can see into those darker areas which the codec threw away at capture. Bad.
So, what this tells us is that it’s very hard to brighten a highly-compressed image, in post. You start getting all kinds of nasty compression artifacts if you brighten too much in post.
My approach to exposure: Overexpose!
So to avoid this problem, I expose almost all of my shots too bright, if I can get away with it. Of course I want to make sure that I don’t clip the highlights, because that would create an even bigger problem. So my technique is: I get the shot looking the way I like it on the screen. And then, I brighten the exposure until the point of clipping highlights, and then back down just a hair.
That way, I’m guaranteed not to have to deal with nasty shadows. I can always darken back to the original exposure. I know that this probably sounds strange, but it really does improve my image quality. It’s hard to demonstrate here, but check out the video for a detailed example.
I did not have a reliable monitor on this shoot. I didn’t even have a sun-shade or eyepiece. So I used the scopes from Magic Lantern for exposure, and they served me very well. (Part Three on Magic Lantern is coming soon!) The Magic Lantern histogram has a really nice feature to alert you when you’re clipping somewhere.
Tips and tricks for increasing dynamic range
I also have a bunch of little tricks that help me preserve my dynamic range and avoid clipped highlights, because while Cinestyle/Prolost Flat definitely do help preserve dynamic range, they’re sometimes not enough.
Stick an object very close to camera to take down a hot spot.
If I have an object in my shot that is too bright to avoid clipping, I’ll often stick an object in the foreground to cover over the bright spot. The blur of the foreground object (a railing, a tree branch, etc) will bleed over onto the brighter object and effectively darken it.
Of course, it will also blur that bright object a bit. But those hotspots are usually in the background anyway, and I don’t mind if they’re a bit blurrier.
Use bokeh to hide hot spots
If there’s a hot-spot in your video, throw it out of focus and it won’t look NEARLY as bad. The bokeh blends it in with the surrounding colors and actually darkens it. If the hot spot is small, the bokeh will completely prevent it from clipping.
And even when the object is so large that the center still clips when out of focus, the clipping will be much less offensive to the eye. Clipping is a loss of detail. When the clipped object is strongly out of focus, we don’t expect to see detail there anyway, so when the detail is lost because of clipping, we don’t really notice. If the object had not clipped, it would have looked practically the same (it would have been a big grey blur instead of a big white blur).
The worst is when you have a clipped object in focus, because we can see the object clearly and we know that there should be detail there, but it’s clipped. Avoid that at all costs.
Here are a few of the shots from the Portrait of Macerata video where I used bokeh to hide clipping:
The circular polarizer rotates on its ring, and it blocks a different angle of reflected light, depending on how you rotate it.
It can be a life-saver for highlights, though of course it won’t work all the time. It only works when the source of the highlight is coming at a different angle from the rest of the light in your shot. I used it mainly for skies in the Portrait of Macerata video. The light coming from the sun and bouncing directly off of the sky is usually coming from a different angle than the rest of the light in your scene, so the polarizer can take down the brightness of the sky, leaving the rest of your shot intact. This works most often early or late in the day.
A note about subject matter
Note that the decision to be super protective of the highlights fit the style of my piece well. If I’m shooting interviews, then I have a single subject that is extremely important in the frame. In that case, I might end up letting the highlights clip a little, because the subject of my interview is MUCH more important. But even then, I would probably try to throw the clipped highlights out of focus if possible. That would probably be feasible.
The Elephant in the Room: raw
“Most of the issues you’ve talked about are because of the codec, right? Wouldn’t it be easier to shoot in raw and forget all of this?”
Yes and no. There are lots of great advantages to raw, but there are downsides as well. Raw cameras are getting cheaper (the BMPCC, for one), but they still cost at least double the price of this camera. The cards are crazy expensive. I shot entirely on SD cards for this shoot for maximum flexibility (remember, all I had was that shoulder bag. I wasn’t carrying a laptop or spare drives around with me), and I shot about 70GB of footage on two SD cards. The same amount of footage in raw would have been 1.53 Terabytes. That’s a LOT of cards. The cheapest supported SD cards for the BMPCC are $86 for 64GB. I’d need 24 of those cards to fit 1.53 TB, so that comes to $2064 just in SD cards! I’d have to be swapping cards left and right. Definitely not my shooting style.
I also relied heavily on the features of Magic Lantern in order to use this camera without lots of accessories like monitors. Magic Lantern is only available for DSLRs.
Well, that wraps up the camera/codec side of things. I’ve got the third video, in-depth on Magic Lantern, already shot. I just need to get that edited, and it’ll be up soon! The fourth video will be all about Post-Production on this film, so send me all your questions! The fifth video will be a tutorial on how to build my motorized slider.