I have been looking at video codecs and it has been revealing, but also a little confusing.
To my mind, I have been thinking of capture codecs and post processing a little like running a shop, something I have experience with.
“Front of house” is a term used to describe what you do at the customer facing direct contact end, so basically the capture point of video. “Back of house” is the running of the business in an administrative and stock handling vein, so it is the post shoot handling of the captured materiel and the pre-shoot prep.
Sometimes you can do most of the work front of house with lighting, composition etc and if done well enough, that is often all you need. If the footage is to be used live, or with little or no processing added, then front of house is all you have.
Back of house is relying on post processing, which requires time and settings that will allow that.
Why?
Because processing of these compressed files is always destructive resulting in files breaking up and loosing visual integrity. You simply cannot shoot to fix it later, but you can give yourself a little wriggle room.
Two things effect this, one more than the other and for a while, I had them a little mixed up.
Codec or the capture type, ranging from IPB/Long-GOP, All-i, to RAW will determine the bit depth and compression. Unless RAW is used the colour profile will determine contrast and saturation applied, which in part effects compression etc, but less so.
Compression makes files smaller and easier to share, but reduce processing depth and require “unpacking” by the computer, which can be a strain. The less compressed the file is, the easier it is on your system to process but the original files are bigger.
I will now throw out a rough guide based on my understanding with a ratio of ten values for front of house:back of house, meaning combinations that require mostly front of house controls applied like lighting, white balance, exposure etc, or back of house like post exposure ISO and white balance, colour grading etc.
You should always employ best practices, but sometimes the choice is made for you.
There are other factors at play of course, like bit depth, specific camera performance etc, but these have less effect overall, so I will assume they will be addressed as suits the user and camera.
As an example, this is how I see my RAW stills; 6:4 meaning nearly equal work at the capture and processing ends, leaning always towards capture as the more important. If I shot JPEG’s it would be more like 8:2, although heavy photoshop users may think it is closer to 5:5.
A RAW stills file usually needs a little coercing, but it does not resist this input and more.
9:1
IBP or All-i codecs"*, with basic camera selected Rec709 colour profiles like Standard or Natural or even more video specific ones like Cine-V. These give you literally what you see on the screen, but with little room to move. Reducing saturation, contrast and shifting white balance are very limited and tend to break up the files quickly, although adding is less restricted, which is why many reduce contrast and saturation. You can give yourself a little more room to move by reducing saturation and contrast, but maybe only a 8.5:1.5 shift is achieved.
For set interviews and less busy subjects with controlled light and time to set the camera correctly, these are fine and are often equal to much higher grade codecs in end results. The reality is, this is where you are headed, so it is possible to get there at the start, as long as you control all of the variables.
I have had great results from my G9 mk1’s in Standard profile, Long-GOP/422/10 bit, as long as I get white balance right in camera and avoid extreme exposure ranges.
8:2
Semi-LOG colour profiles like Flat or Lite-LOG allow more processing of colour and contrast, but do not effect compression. You have some post control, but errors in white balance etc are still not easily fixed.
For a while, this was my maximum quality level, being easier to use than full LOG which requires extra exposure awareness, but more flexible than a standard colour profile.
7:3
Moving up to LOG or LOG-like colour profiles adds a relatively increased amount of processing options. Full LOG profiles are heavily softened, allowing processing to add colour, contrast and exposure information from a very forgiving base, but also require exposure and white balance awareness and often need viewing Lut’s applied.
I avoided LOG for a while when it was available, but the reality is, if you are limited in codec choices, it is the best option for serious video creation, as much for it’s support structure as anything else. Learn as much as you can about properly exposing LOG.
6:4
Next we change codec to something like Apple ProRes which delivers less quality loss with less compression. Colours etc are still baked in, so up front choices are important, but overall quality is potentially higher and processing relatively easier.
The colour profile you choose is still important here as what goes out to the capture device is ProRes with a colour profile like LOG applied, so in camera settings are important. Personally I am interested in knowing if a standard profile like Natural and ProRes 4:2:2 HQ would make a difference.
The files below were shot with the wrong WB setting and exposed for the open side of the face. It pushed and pulled well enough, but the test was against B-Raw and in this circumstance, with my limited skills, there was no comparison.
ProRes 422 may be an option as a better quality choice than those above, but 422HQ has not impressed me with it’s strong contrast and less malleable properties in post than B-Raw and the size is considerable, even in 1080.
5:5 RAW formats in video are a complicated thing, but let’s assume they mean much the same as they do for stills, which are effectively un-compressed, unprocessed files.
Two things happen here. The lack of compression (althought here are sub-choices within the format) means files can be large, stressing capture devices and storage, but as there is no compression at the outward end, computers can handle these more easily (there no compression to “unpack”), as long as the file size does not bother them.
The reason I have applied an even higher ratio to this than stills is the reality that moving footage often requires more work, even if it is harder and less efficient. With stills, it is totally realistic to assume a well exposed RAW capture may only need a little saturation or contrast to get sorted and the reality is, most processing platforms apply a small amount of pre-set information, while ironically, video footage, which can often have a Lut applied (a pre-set look up table of settings), often needs that and more, which being connected and sequential is often harder.
The file below was a pleasure to grade. I bought it up to bright and cheery like the last file, made it more or less contrasty and adjusted white balance easily. Shadow detail was bought back up (deep blue using false colour, so I would have assumed completely lost), highlights showed no sign of blowing out and my wife’s hair, which broke up a bit as she spoke in the ProRes files, was clean and sharp.
The best bit is, the C4k constant quality Q5 RAW output, was only slightly bigger than the 1080 ProResHQ footage.
B-Raw in Q5 constant quality seems to be a sweet spot and is recommended by Black magic for “TV serial content”. It processes like RAW stills, without the oddness of LOG profiles or the still present limits of ProRes.
The reality of these things is clear to me. Any capture format can work, but the more flexibility you want, need or are forced to use, the more information you need to capture and retain.
My two baselines are ProRes/Flat for simple and controlled jobs, ProRes/Log (or just LOG) when I need more latitude and either cannot use or do not have a BMVA to spare, but may need to match it, and B-Raw/Q5 when I can, but limited to a single unit at the moment.
I would like to like ProRes more, but B-Raw is clearly stronger, more flexible and intuitive and lighter on my system (more hard drives are cheaper than new hardware).
*IPB such as Long-GOP are a combination of bespoke files supported by extrapolated files, the most common codec available in consumer cameras or All-i using clean files per frame,