Green Screen & Chroma Keying: Lighting, Spill & Mattes

1. Green Screen vs. Blue Screen: The Physics Choice

Choosing between chroma green and chroma blue depends on digital camera sensor architecture and wardrobe:

• Why Green Dominates Digital Cameras: Modern digital cinema sensors use a Bayer Color Filter Array (CFA) with twice as many green photosites as red or blue (RGGB pattern). The green channel delivers the highest spatial resolution, cleanest luminance signal, and lowest electronic noise, making clean matte extraction effortless. Green also requires significantly less lighting power to illuminate.
• When Blue Screen Is Superior: Blue light reflects far less visible light back onto human skin, making blue screen ideal for blonde actors with fine wispy hair, reflective glass/metallic props, or scenes featuring green costumes and military uniforms.

2. Lighting the Cyclorama: The Exposure Rules

The number one mistake on visual effects sets is over-lighting the green screen backdrop:

3. The Distance Rule & Negative Fill Spill Suppression

Light intensity falls off with the square of the distance (the Inverse Square Law):

Position actors at least 6 to 10 feet in front of the green screen. This physical air gap dramatically reduces green bounce spill onto the actor's shoulders and neck.

To eliminate residual edge contamination, gaffers rig a subtle Magenta Rim / Hair Light (Minus Green gel) behind the subject. Because magenta is the complementary opposite color of green on the color wheel, the backlight neutralizes green edge bounce into neutral white.

4. Codec Bit-Depth & Chroma Subsampling

Never record chroma key shots in highly compressed 8-bit 4:2:0 consumer codecs:

• 8-bit 4:2:0 H.264: Color information is sampled at only one-quarter the resolution of the black-and-white luminance image. Keying algorithms encounter severe pixelated staircase stairstepping around fine hair and motion blur.
• 10-bit 4:2:2 & 12-bit ProRes 4444 / RAW: Retains full color gradient precision (1,024 levels per channel instead of 256), allowing keyers to isolate subtle semi-transparent motion blur and smoke with smooth sub-pixel alpha gradients.

5. Advanced Compositing: The Dual-Matte Workflow

Professional compositors (in DaVinci Resolve Fusion, Nuke, or After Effects Keylight) never pull a single global key. They split the extraction into two complementary mattes:

1. The Core Matte (Holdout Matte): A high-contrast, aggressive key with clamped black and white thresholds that fills the interior of the actor's body solid white (100% opacity), preventing transparent holes in shirts.
2. The Edge Matte (Detail Matte): A delicate, gentle key tuned exclusively to preserve fine semi-transparent strands of hair, motion blur, and glasses edges.
3. Combine & Despill: Merge the Core and Edge mattes together, apply an optical despill node to restore true color to green-tinted fringe pixels, and composite over the CGI background plate.

Tested: chroma subsampling, backdrop evenness and exposure, with FFmpeg's chromakey

We tested two of the page's claims on 2026-09-29 with FFmpeg 8.1.2 and numpy, on generated green screen frames whose true alpha is known. Each frame was encoded with x264, keyed with FFmpeg's chromakey filter after upsampling the chroma to full resolution, and the key was compared with the true alpha. FFmpeg's chromakey is a simple keyer that works mostly on chroma and gives a nearly hard matte, so this shows how the page's claims play out with that keyer, not with Keylight, Fusion or Nuke. The frames are synthetic, with no real hair, motion blur or spill, and nothing was looked at.

Section 4 holds in direction and is milder than "severe" here. Chroma subsampling to 4:2:0 keeps one colour sample for every four brightness samples, as the page says. On a smooth skin colour edge the average alpha error was the same in every format, 0.023 at 4:4:4 and 0.024 at 4:2:0, and the only difference was that 4:2:0 and 4:2:2 got 6 and 7 pixels out of 5413 in the edge band badly wrong, against none for 4:4:4. On fine dark hair the cost was larger. With strands two pixels wide, 4:2:0 lost 36 percent of the strand pixels against 25 percent at 4:4:4, and 4:2:2 sat between at 32. Strands one pixel wide were mostly lost in every format, 87 percent even at 4:4:4, which is a limit of this keyer.

The bit depth claim did not show up. 4:2:0 at 8 bits and at 10 bits lost the same share of hair, 36 and 36 percent, which suggests that the sampling, and not the number of levels, was what limited the key here. The page's case for 4:2:2 and 4:4:4 rests on the sampling. Its case for 10 bits, smoother gradients in smoke and motion blur, was not tested, since this keyer returned a hard matte for those.

Section 2's half stop rule is a safe margin, not a cliff, for this keyer. With a tight key, a screen that varied by half a stop either side of its middle left 0.1 percent of the background partly opaque, and one that varied by a stop left 21 percent. With a looser setting, it left 0.0 percent at a stop and a half either way. So how even the screen must be depends on how tightly the key is set, and a loose key trades that for a greater risk to the subject's own colours, which this test did not measure.

Exposure was more forgiving on the bright side than the dark one, until it was not. A screen a stop darker left 100 percent of the background partly opaque, a stop brighter was keyed cleanly, and two stops brighter clipped in the frame and was not keyed at all, leaving 100 percent of it opaque. That supports the page's warning against over-lighting for the clipping reason. Its other reasons, green bounce onto the subject and hair turning white, need a lit subject and were not tested. The exposure figures depend on where our green sat in the range, at 177 of 255 in the green channel.

Key a green screen in FFmpeg with a similarity and blend like the tight setting used here: ffmpeg -i in.mov -vf "format=yuva444p,chromakey=color=0x00B140:similarity=0.10:blend=0.0" -c:v prores_ks -profile:v 4444 out.mov

Validated. Measured 2026-09-29 with FFmpeg 8.1.2 (x264 and chromakey) and numpy, on synthetic frames. FFmpeg's chromakey is not a production keyer, the strands and edges are generated, and the results would differ with real footage, motion blur, spill and other keyers. We did not test despill, the dual matte workflow, blue against green, the magenta backlight, the distance rule, the 1 stop below the subject rule, sensors or how any key looks. The command is the key that was measured, with the output codec changed to one that carries alpha.

Working Vocabulary

Key VFX compositing and chroma keying terms introduced in this guide:

• Chroma Key. A visual effects post-production technique for compositing two video streams together based on color hues (typically green or blue).
• Green Spill. Reflected green light bouncing off a cyclorama onto the subject, contaminating skin tones and clothing edges.
• Despill. A mathematical color transformation that isolates green fringe pixels along subject edges and neutralizes them into matching skin or background tones.
• Core Matte. A dense solid alpha mask used to protect the interior opaque areas of a subject from becoming semi-transparent during keying.
• Edge Matte. A delicate soft alpha mask dedicated to preserving fine details such as hair strands, motion blur, and fabric fibers.
• Bayer CFA. The microscopic color filter mosaic on digital sensors with an RGGB layout containing twice as many green-sensitive pixels as red or blue.

Sources & Further Reading

• 3-Point Lighting & Cinematic Contrast Ratios
• Blender 3D Camera Tracking & Live-Action VFX
• ACES vs. DaVinci YRGB Color Management Guide
• Video Frame Rates & 180-Degree Shutter Angle Rule