May 18, 2017 at 8:42PM, Edited May 18, 8:50PM


Can we have a nerdy discussion about video bit rate vs resolution?

In the midst of talking over an upcoming shoot today where we'll be using a Sony A7SII recording Slog out to an Atomos Ninja Flame (w/ Rokinon cine 24/35mm f/1.4s and Zeiss 50/85mm f/1.4s ), my partner (who DPs for our production company) and I started a conversation about bit rate vs. resolution and both of us realized we were both a little hazy on the relationship. Is the higher bit rate getting us the detail or is the resolution? If it's both overlapping, how would we prioritize for different sorts of shots? If, for example, we wanted a get a wide master at 4k 24p with the 35mm (i think 100mps? don't have the menu open here) are we getting more detail in general in the wide than when we do a 1080p 60p medium shot (at 50 or 60mbps i think?), if the distances were accounted for and we punched in on the master? I'm sure it totally depends on the content- one actor in a hazed out, heavily gelled LED light flood, cyc wall situation- but neither of us were sure. I guess we just aren't totally sure how bit rate plays against resolution where detail is concerned - if the new gh5 can do 10-bit 422 4K at 400 mbps VS our 10 (with 2 empty, due to a7sII HDMI limit) bit 422 prores 4k at 100mbps, what is it we're NOT getting? is it 300mbs of detail in MOVEMENT per second, or summed between frames? if 300mbps divided into the FRAMES per (24/60 etc) is it fine detail in hair and fabric, consistency like in motion blur? I just honestly just don't know exactly what these number/metrics mean in IMAGES. Any insight would be helpful - everything matters when it comes to telling the story well. I want to at least understand the technicalities well enough to be able to make necessary considerations regarding movement, detail, color, light etc all having their best chance to play their role- thanks in advance, sorry if that was overly convoluted -_-


If you record internally with your camera, then you should also factor in the CODEC you are using. The XAVC CODEC is very efficient at packing detail into a smaller file compared to something like the Apple ProRes CODEC.

Also some cameras have better implementations of a CODEC, so 100 Mbps MP4 on one camera might be better than 100 Mbps MP4 on another camera.

May 19, 2017 at 3:53AM

Guy McLoughlin
Video Producer

Resolution is a somewhat ambiguous term, but in the video world it usually means line pairs per mm (or something that equates to that). If your subject is a black-and-white resolution chart, "resolution" tells you how many black/white or white/black transitions you can accurately measure in a given area. But...a perfect score here may not do so well when it comes to imaging blue sky: is it white or black?

Now we come to the second dimension, which is color depth. Without sufficient color depth, you get color banding, which can sometimes be fixed in post-production with dithering.

There's a third dimension, which is the relationship between luminance resolution and chroma resolution. The human vision system can fill in missing color detail and/or correct slightly wrong color information. But when colors are used as data (such as for green screen/key operations), low chroma resolution translates to low spatial resolution at the edges of keys, which is a problem.

As you rightly point out, video compression algorithms try to be smart, allocating bits to things the eye will pay most attention to and using minimal bits to things the eye will forgive. But with a finite number of bits per second, there will always be a compromise between (1) how many actually resolvable features there are, (2) how accurate are the colors of the high-detail features, (3) how accurate are the colors of low-detail features, and (4) how accurate are the colors when defining "key" areas (if you are keying).

What you are not getting with a7sII HDMI output is real 10-bit 422 output, but rather 8-bit 422 output. Adding an extra 2 bits means quadrupling the data rate, which is...400 mbps. 8 bit output means more banding, which is probably more of a problem for large-scale features, like blue skies or shadow areas. Well-lit hair or fabric will probably look just as busy at 100mbps as 400mbps.

Your wide vs. medium vs. narrow shot question is a red herring. Wide shots may have more high-detail than narrow shots (lots of small leaves vs. a few large ones) or they may have less (lots of blue sky vs. the feathers of an eagle as it flies in the sky). For any given shot the questions are: how much fine detail is there (and does it blow out to noise because there are not enough bps)? how many large gradients are there, and are they banding in a way that's distracting? if you need to key, are your key edges sharp enough to give a good result based on the nature of the subject and its background?

May 19, 2017 at 4:24AM


Thank you, that spare (empty -_-) 2bits = quadrupling the bit rate info was eye opening

May 19, 2017 at 8:09PM, Edited May 19, 8:09PM

D. Scales
director / producer

Your Comment