Color correction + grading workflow concept

[DamienMTL]

Hi,

I come from a sound engineering/mixer background and currently improving my video skills. And I shoot a lot of photos, with neutrality and balance in mind before providing a personal flavor. I always use color charts to start with...

Coming from this sound processing background, we try to correct each audio tracks to  get a natural and balanced sound before giving it a special tone/taste.

I conceive video as audio or photo, maybe I'm wrong.

I had a GH2 and bought a GH4 and everybody is talking about the different color profiles (cinelikeD etc...) some being better than others for color grading.

Let's say you start with a neutral/natural look as close to reality, (red is red, green is green,... black is black) then tweak it to taste for a dreamy or punchy or gloomy look...

For that I shoot some gretag macbeth charts, like the X-rite colorchecker passport, with the cinelikeD or natural profile. then I use the plugin "MBR Color Corrector 2" which can produce a shot by shot correction to natural, neutral and balanced image. DaVinci resolve does it too. And it's perfect for matching scenes and different cameras, which could be a nightmare. It's simple, quick and easy, and always spot on for every color.

Then, I add a personal grading to my taste. Here comes the arty job.

Is the process wrong somewhere?

Am I losing the flavor of the lens or camera or the tone of the camera profile?

If everybody is shooting videos with the mandatory grading step in mind, why not shooting "neutral" vs already flavored by a color profile like cinelike D?

Do I lose data-information or is it the appropriate way to do?

If so, I don't see a lot of pros shooting with 24 color patch targets and why? I see people with grey cards for white balance and that's it. Is my conception of color rendering and grading wrong, am I missing something?

thanks for your thoughts about all that.

[Sekhar]

Charts are great for stills, but in video the image is continuously changing, with usually varying lighting. E.g., if a camera pans, are you going to put your chart in beginning of the pan or the end? Unless your lighting is relatively uniform, charts are kind of useless IMO. Obviously you'd typically want to color balance while recording, but it'll almost always be necessary to eye-ball the shot as you correct in post, which is really simple anyway unless the recording format makes it hard. Which brings us to the necessity to record in a way that allows for flexibility in grading (RAW, log, 10/12 bit, etc.).

[Axel]

Is the process wrong somewhere?

​No. It's thorough and reasonable.

If everybody is shooting videos with the mandatory grading step in mind, why not shooting "neutral" vs already flavored by a color profile like cinelike D?

​A misunderstandig. Profiles with cine in their name usually refer to very neutral values. They are also called flat. That means they have a very flat gamma curve applied. For 8-bit this results in (hypothetically) 256 luminance values almost evenly spread. By grading this flat video, you can apply your own curve in post ...

*BUT*

There is a reason that profiles like landscape or portrait exist. I elaborate on these two examples. With a landscape, you have the sky in the image. It is (really!) a few thousand times as bright as the rocks in the shadow. To depict this sky in it's glory, you need quite a big portion of the 256 on the bright side of the spectrum. But you also like to see the rocks. So many values get spared for the lower mids. The least values remain for the mids. Someone walks by. His/her face may be of natural 'porange' (pink-orange, term coined by Magic-Bullet guru Stu Maschwitz), given you had roughly the right WB and exposure, but it looks like a colorized pancake nonetheless. Because for natural and rich skin tones, you also need well quantized mids, and those weren't baked in the 8-bit video. In portait mode, you may have a sky that looks like watercolored, but the person will look good.

Think of it like audio frequencies. You have a camcorder stereo mic and a person in the street ten feet away, saying something important. Bad choice. You'd try and use the microphone that prefers voice frequencies, a lav perhaps. 

Only with raw (where nothing but the bare signals are stored) you can hope to change colors completely. You have better chances with 10-bit (Shogun?), but generally it's a good idea to record neutral and many colors. Avoid clipping, avoid underexposure, choose a profile that fits, expose skin tones @ 50% in the histogram, if they are important.