Now, H.265 will obviously offer better quality (for the same size) compared to H.264, so this may also be a factor.

One way to (at least partially) counter the lack of color info with chroma subsampling is shooting higher resolutions. For example, shooting 4K with 2K end result in mind will deliver a bit denser final image because of oversampling. Actually, this is probably the only decent way of shooting flat images on a consumer or prosumer level equipment.

I’d like to toss in a few additional bits (no pun intended.) The compression algorithm used in all modern consumer cameras and camcorders for video is called MPEG-4 H.264 AVC. It partly works in the same general way that JPEG works: discarding detail that human eyes aren’t likely to notice. Unfortunately the infamous so-called “YouTube compression” that is blamed for significantly reducing visual quality of some videos is the exact same compression used by your consumer-grade DSLR or camcorder. The thing about AVC is that it has a ton of “knobs” that can be tuned to change how it works; some “knobs” make the resulting video incompatible with older devices; some would boost quality but also increase the time needed to compress beyond what can be done in 1/30 of a second in a cheap ASIC chip so they don’t tend to be used by consumer cameras at all.

Another thing that people don’t realize is that consumer AVC video is recorded in a “chroma subsampling” pixel format called 4:2:0, where 100% of pixels have their luminance (brightness) information recorded but the COLOR INFORMATION (chrominance) resolution is HALVED IN EACH DIRECTION. That means color is only sampled once for every 2×2 block of pixels. You can see the problems caused by this discarding of 3/4 of the color information when you take SD footage and try to use a green screen and chroma key; the boundary between the subject and the green screen gets “blurred together” by the subsampling and it’s not so easy to get a clean chroma key. HD footage has the same problem but 1080p has more than double the pixels in each axis which sort of cancels out the halving of color information.

This all matters a lot when you change from “looks good in-camera” picture styles to these flat styles that supposedly make things look better. Anyone who has watched a heavily compressed DVD rip has seen MPEG-4 macroblock banding in dark frames and other low-contrast areas. Some advanced AVC encoders have a knob called “psychovisual rate distortion” that can be used to push more of the available encoding bits into these low-contrast areas, but consumer cameras can’t really optimize this for all situations, so they make up for it with higher overall bit rates (think about a 24 Mbps consumer video to a 1 Mbps DVD rip; even with 6x the info in 1080p vs 480p, the 24x bit rate can encode 4x more of the video info that would otherwise be discarded.)

Despite higher bit rates, the visual problems are still there, and pushing up low-contrast areas in post will not only reveal more noise, it’ll also make MPEG-4 macroblock banding much more obvious. This comes full circle when you combine “shoot flat (low-contrast)” and “MPEG-4 causes banding in lower contrast areas.” By shooting flat on a camera using an 8-bit 4:2:0 MPEG-4 AVC output codec, you’re making that consumer-grade camera discard A LOT MORE visual information than it would if you had just used a “neutral” style. The histogram breaking up is going to be a problem with any extremely flat picture style, but the last thing you need is to introduce barely visual banding artifacts on purpose, THEN boost their contrast AND break up the histogram to cause even worse banding.

If you have a fancier camera that does 10-bit or 4;2:2 encoding then it might be less of an issue; if your camera outputs huge raw video files or dumps to an uncompressed video recording box, you can do this flat thing all day long, but if you’re reading this then you don’t have $15,000+ for such a fancy pile of gear. Know the limitations of what you have and test the extremes in post of things like custom picture styles before you put something important in that format and find out the hard way that you can’t use the footage the way you had intended. Work smarter, not harder!

That said, as long as it works ok for you, then it is fine. :) Requirements vary from one person to another, so nothing is cast in stone.

Underexposure is always bad. It was generally bad with film, and it is bad with digital cameras. There is too much noise down there.

But I don’t think Cinestyle has more detail in the shadows than the others. It kinda appears this way cause of the lifted black level. I wouldn’t really bring the shadows up in post with any flat style, they’ve already been pushed enough by the PS. If they need more pushing, then most likely the exposure wasn’t right in the first place.
And I’d still use Neutral/Faithful when a serious amount of grading is expected – Cinestyle just doesn’t have the fine shades in skin. Its softness can be a positive for some applications, though.

The popularity of brought down saturation probably originated from Stu’s preferences back when he proposed Neutral with -4 Contrast and -2 Saturation. Seems like he likes slightly desaturated images. The problem with picture style desaturation is similar to picture style low contrast: you will get banding in chroma when you increase it in post, the same way you get banding in luma with returning contrast. It is generally better to remove saturation in post, rather than add it back. I also find that Sat at -1 works well for most purposes.

Thx for your perspective. :)

I will just shoot with Faithful/Neutral if I don’t need the shadows really open though. All flat styles seem to introduce some minibanding. It is not quite clear exactly on what stage, in what color space (and with what data precision) they are being applied during the in-camera processing.

I actually prefer Faithful over Neutral for its blues. Not that fond of its skin tones though: Faithful can push skin hues too pink.
Flaat_10 is pretty nice and with a very predictable distribution which makes it easy to grade.

Sometimes it is better to just leave blacks of whites where they fall, even if they clip and not worry too much about zebra alerts. This is especially true if you are going for a descently contrasty look anyway. No need to have precious bits taken from the mids if you are gonna press the shadows back down in post.

There is also the pain of matching the range through different shots in a sequence if they were shot with different styles (even with matching colorimetry, like Neutral and Flaat_10n, for example). While it is perfectly possible (and in fact – optimal – image-wise) to get every shot with the perfect possible DR distribution, matching these together and making sure important tones fall at the same values in the end is a time consuming task.

Bart