SGO Mistika provides real-time, UHD and 8K resolution workflows.
This third and final article of the series about color grading examines more the process with an eye to creating an efficient workflow. The goal is to properly prepare the content before it is moved onto an editing platform. Preparation becomes even more important when the editor uses one brand of NLE and a different solution for grading.
Part 1 of this 3-part series explained why Media Composer | First and some other NLEs cannot support round-tripping—the usual way of color-grading with DaVinci Resolve in combination with an NLE. The first Part also covered the process of setting up Resolve for a grading project, thus it also serves as a general Introduction DaVinci Resolve.
Part 2 covered the import of different types of media for grading and/or for editing. Part 3 will focus on rapid color-grading using Resolve.
Before getting into the details of the grading process, there are several topics to be considered. First, there is the difference between high-contrast and low-contrast scenes. Second, rapid grading depends on reasonable assumptions made about the nature of very dark and very bright areas within a scene.
Figure 1 presents the 11-segment Ansel Adams Zone System adapted to Resolve’s brightness levels ranging from binary “0” to “1023.”
During import, digital video data range from “16” to “235” (8-bits) or from “64” to “940/960” (10-bits). This 100% range is “broadcast legal.” However, a 108% range, is possible if the camera and codec record digital data up to “255” or to “1016/1023”. (Most cameras and camcorders capture data that range from “16” to “255.”) This wider range is acceptable for presenting on a computer or for streaming.
Imported data are converted within Resolve to YRGB 32-bit floating-point values. Before being displayed on Resolve’s 0 to 1023 display scale, 8-, 10-, and 12-bit values are converted from floating-point data, to integer data that range from 0 to 1023.
During grading when values are pushed below 0 and/or above 1023, because they are floating-point values, they are preserved despite Resolve no longer displaying them.
For this reason, when I see data crammed against 0 or 1023, I always dial Lift (Shadows) up and/or Gain (Highlights) down so I can see what data might be obscured below or above the lower and upper limits of Resolve’s scopes.
As shown by Figure 2, a high-contrast scene may have signal levels that cover the full range from low 0 to high 1023. (The scope images herein are screenshots from a newly released digital B&W film.)
Figure 3 presents the scope display of a dark, low-contrast scene. Both Parade and Waveform displays are shown. Note that dark, low-contrast video continues to use most of the signal range. Also note, a cinematographer will almost always include a very bright object, ≈896 (90%), within a scene to establish contrast. (Think, Touch of Evil directed by Orsen Wells.)
Figure 4 presents the scope display of a medium-brightness, low-contrast scene. Note that most of the signal range is used for low-contrast video. Also note, there are highlights above ≈896 and shadows down to ≈64 to create adequate contrast.
Figure 5 presents the scope display of a bright, low-contrast scene. Note that low-contrast video uses much of the signal range. Also note, a cinematographer will almost always include a dark object, ≈320 (30%), within a scene to establish contrast.
Color-grading involves making an incredible number of often very subtle adjustments. Although I’ve rarely had Resolve crash, it only makes sense to enable its auto-save feature. See Figure 6. Issue the Davinci Resolve > Preferences… > User > Project Save and Load command and check Live Save.
Figure 6: Enable Auto-save Project. Click to enlarge.
Figure 7 shows the Color room. after engaging the scopes by issuing the Workspace > Video Scopes > On + 2 Up command. Then select (RGB) Parade and Vectorscope and then position the scope window as shown.
On the Vectorscope, select Show Skin Tone Indicator (line) and Show Graticule.
Select Color Wheels and Primaries Bars (red) because they function intuitively with the RGB Parade scope. See Figure 8. The green box marks the Red, Green, and Blue sliders for Lift (shadow), Gamma (mid-tones), and Gain (highlights).
The white box marks the Lift, Gamma, and Gain dials that equally raise and lower the RGB levels.
The yellow box highlights the Saturation and Hue controls.
Figure 9 shows the RGB Parade scope next to the Primary Bars.
To move the bottom (cyan circles) of a Red, Green, Blue bar in the RGB Parade scope higher or lower, drag the upper and/or lower edges (red circle) of Red Lift, Green Lift, Blue Lift slider(s) up or down.
To move the top (white circles) of a Red, Green, Blue bar in the RGB Parade scope higher or lower, drag the upper edges (green circle) of Red Gain, Green Gain, Blue Gain slider(s) up or down. You cannot drag the lower edges of the three Gain sliders.
To move the middle (yellow circle) of a Red, Green, Blue bar in the RGB Parade scope higher or lower, drag the upper and/or lower edges (magenta circle) of Red Lift, Green Lift, Blue Gamma slider(s) up or down.
It’s now time to consider the assumption that can make grading fast. The assumption: in most scenes where the camera is set to correctly match the illuminating color temperature, the darkest and brightest objects are captured as a monotone grayscale.
Simply put, the darkest portions of a scene will range from dark-gray to black. Likewise, the brightest portions of a scene will range from white to clipping.
Obviously, there are exceptions such a scene in a silent running submarine as shown by Figure 10, or even a cloudless, bright blue-sky day.
Figure 10: Fluorescent Lights Turned-off, Red 60Hz Incandescent Lamps Illuminate a Submarine. Click to enlarge.
Set Lift: Glance at the Viewer to confirm that the darkest object(s) should have no color. In Figure 11 you can see the garage entrance is very dark gray.
Drag the bottom of the Red Lift, Green Lift, Blue Lift slider(s) (see Figure 8) until the lowest RGB Parade scope signals are all at zero. (Figure 12.) If you find dark grays are crushed to black, set all of the lowest RGB Parade scope signals to 64. (To obtain an even lower contrast image, you can set all to the lowest RGB Parade scope signals to 128.)
Set Gain: Glance at the Viewer to confirm that the brightest object(s) should have no color. Figure 13 has multiple very bright buildings all of which are white or very near clipping.
Drag the top of the Red Gain, Green Gain, Blue Gain sliders (see Figure 8)until the highest RGB Parade scope signals are all at 1023. (Figure 13.)
When the brightest object(s) should have a color, such as here where the sky is bright blue, switch to Primaries Wheels and drag the Gain Puck toward the desired color—in this case Blue. Then use the Gain Dial to place peak Blue at 1023. See Figure 14.
Set Saturation: Adjust Saturation to the desired strength using the “Sat” control. (Figure 15.)
Unify mid-tone coloration with the coloration that was set using Lift and Gain sliders. Do this by adjusting the Red Gamma, Green Gamma, Blue Gamma sliders. (Figure 16.)
For example, the portion of a face that is in bright sunlight should “match” the portion of the face in shadow. (You may need to re-adjust Saturation.) You want one person’s face under different amounts of illumination—not two different faces.
Set Gamma: Dial Saturation (Figure 15) to zero and then, using the resulting monochrome image, adjust Gamma, using the Gamma Dial until the entire range from Black-to-White looks “natural.” (Figure 17.)
Set Hue: When a person is within the frame, adjust the Hue control (Figure 15) until the signal representing flesh aligns with the vectorscope’s Skin Tone Line. This line is valid within a few degrees for all skin tones. (You may need to re-adjust Saturation.) See Figure 18.
The center of each (Red, Magenta, Blue, Cyan, Green, and Yellow) vectorscope “box” represents a saturation of 75%. Female Caucasian skin is about 40% saturated while male skin is about 35% saturated. As shown by Figure 18, saturation was adjusted until the tip of her skin vector was at ≈40%.
Next, set the desired skin brightness while looking at the RGB Parade scope. Caucasian female skin brightness ranges from 50-75%, while male skin brightness ranges from 45-65%. Figure 18 shows the RGB Parade scope after Gamma and Gain Dials were set so mid-tone Green was ≈768 (75%).
Resetting Lift and Gain: The preceding adjustments likely will have made changes to a scene’s lowest and highest signal levels. Therefore, drag the Lift Dial so the current lowest value touches zero. Then, drag the Gain Dial so the current highest value touches 1023.
Or, maybe not. Looking back at Figure 18 there are two types of “highest signal.” One type are broad signal areas. The other is a series of tiny peaks on the Blue waveform that extend a bit higher. (However, spikes can be quite tall.) So, do we place the tallest spike at 1023 or place the broader body of the signals at 1023 and ignore the spikes?
One solution is to place signal body at 960—midway between 896 and 1023—because it is the highest legal signal. Then the range to 1023 is reserved for spikes.
The other solution is to decide if the spikes represent specular highlights, for example, bright reflections from a chrome bumper. These reflections may be clipped since they have absolutely no detail and may even have a slight tint. Because they contain no useful detail they can be ignored while ordinary bright white objects are pushed to 1023.
The goal of this first color-grading pass is to create a series of clips each of which is technically correct and whose appearance is not flawed. However, even if each shot looks correct, very likely you will need make a second pass to color-match shots that will be in the same scene.
The color matching pass need not, and perhaps should not, be done within Resolve. When you have assembled the clips from Resolve in your NLE’s Timeline, you can use your NLE’s color corrector to match shots. Here too you must keep each shot within signal level specifications by watching your NLE’s waveform scope.
Resolve has many more color-grading tools including keyframed color settings that enables you to, for example, color correct a pan. Power windows is another very powerful function.
Power windows can dynamically track objects, such as the green Triumph Spitfire shown in Figure 19. Then, using two nodes, one can apply different grades to the two windows. The car is minimally graded, while the way too greenish Yugoslavian road is graded to grayish gravel. Once you’ve got the basics of grading down, these more sophisticated techniques are not hard to learn.
To complete this three-part series, here are the links to Parts 1 and 2 of the tutorial series.
Steve Mullen is a video consultant and writer with a wealth of experience in video editing, grading and camera technology. His other articles on The Broadcast Bridge can be located via the home search function. Search for "Steve Mullen".
Related Editorial Content
It's unlikely any editor is unaware of the release of Resolve 15. It’s equally unlikely many editors, especially those who use Media Composer, haven’t thought of switching to DaVinci Resolve. Switching, however, isn’t as simple as it might seem.…
Part one of this three-part series explained why Media Composer | First, and some other NLEs cannot support round-tripping, the usual way of color grading with DaVinci Resolve in combination with an NLE (such as Media Composer or FCP X). Part 1…
This third and final article of the series about color grading examines more the process with an eye to creating an efficient workflow. The goal is to properly prepare the content before it is moved onto an editing platform. Preparation…
The Canon EOS C300 Mark II Digital Cinema Camera is feature rich, offering a re-designed imaging system with a 8.85 Megapixel super 35mm 16:9 CMOS image sensor that supports a maximum resolution of 4096 x 2160 pixels and a high-speed readout for reduced rolling-shutter…