How to Film in LOG: A Beginner's Guide

Excellent feedback below... and in most visual arenas, I agree. Get it right in the camera. 40 years of doing so... BUT, I am now shooting alot more Architectural Interiors, and admittedly, not highly budgeted to light them. Gymbal walk throughs etc. My interest in LOG is based on needing to hold the blown out windows. By "squeezing"the values in LOG, will I not have the opportunity to hold the 6 + stop difference between the ambient shadows in the room, and the unknown ( due to sun/shade etc ) brightness of the windows? I'm not shooting a hi end production, but rather looking for a way to hold the values in the old Ansel Admas grey scale. Any thoughts are very welcome. Thank you.

The best way to shoot in log is: don't. Simple math says why not: you're taking a limited range of values (usually with 12 or more bits of depth per channel, representing 4096 levels), scaling them to a different and smaller set of values (8 or 10 bits of depth, or 256-1024 values) that's drastically different from the set used by the 8-bit (256-value) final product. In the case of 8-bit footage (as most Sony cameras and almost all sub-$1000 video-capable cameras and bodies put out), the problems become obvious quickly, because the mismatch between raw footage and final output values can be quite large, meaning that by shooting in log, you've quite literally just thrown away a lot of crucial image data. With 10 bits, you have quadruple the precision available, so the "compression" of the log curve can't "crush" the lower bits of precision so easily since there are more of them, making a shift back to normal more practically possible. Unfortunately, this only avoids most of the damage; you still have burdened yourself with a more complex workflow and the difficulty of converting back to normal. In all cases, the AVC or HEVC compression used on all cameras today will exacerbate the problem due to the artifacts caused by macroblock compression, and chroma subsampling has also already discarded between 50% and 75% of the color information even before the compression further mangles everything. Shooting log is setting up for failure, and people don't produce good products BECAUSE of log, they do it IN SPITE of log.

The math behind this stuff is why midtones (especially skin tones) look like garbage when shot in log or "flat" profiles and then get pulled back to a more normal image. This is a very common problem that newbies especially face: people will tell them how great log is, with the mythical promise of "more latittude in post," then when they've committed a shoot to log footage and can't simply re-shoot it, they persist in trying to repair their plastic skin tones, ugly midtones, and weird color and jagged-edge artifacts. The pros use log, after all, so surely it's their fault that they're failing! The truth is that log is a good idea in theory and a bad idea in practice.

The correct thing to do is always to get it right in-camera and worry more about what you're putting in front of that camera than what magic setting will make your $5 production somehow look amazing. You can't trick the camera into being a more powerful camera with clever camera settings. 8-bit 4:2:0 AVC-compressed data doesn't care what the professionals told you; it sees the much lower contrast of log curve footage as a giant field of opportunities to discard the subtle gradient detail in a lot more low-contrast areas, so a progression from 0,1,2,3 might become 0,0,2,2 because that's how lossy video compression works: it throws out these little things you won't notice...only you do notice them, because you have to stretch the data back out (particularly in the midtones) for it to be usable, and that 0,0,2,2 sequence might not be 0,0,4,4 instead of something like 0,1,3,4. BOOM! Now you have little blocky edge artifacts in your image and smooth colors like the curved skin of a human face look like a bunch of flattened bands instead.

Even worse, all that extra (noisy) shadow data and extra highlight data that you compromised your midtones to record is just getting compressed right back, so you threw out your midtones to get more highlight and shadow data, and you threw that extra data out too. It's ridiculous. I don't understand why anyone does this. I have never once seen an example of corrected log profile footage and corrected (to visually match) normal/neutral profile footage where the log footage looked better. Even worse, people tend to not sufficiently restore the contrast to log footage because of the aforementioned problems with artifacts and loss of transition details, so corrected log footage video all too often looks like low-contrast, washed-out garbage. Alternatively, recognizing the problem, some log wanderers go the opposite route and over-correct, cranking up saturation and contrast and applying effects to make the problems caused by log footage look like an intentional, stylistic choice, often followed by them making a YouTube video bragging about it and selling their LUTs for $20.

tl;dr: don't shoot log or flat, you're ruining your footage. For more discussion, there's an article "YouTube video experts don’t understand why flat/log footage on 8-bit cameras is a bad idea" out there with a more visual explanation.

thanks!!!