Part 9: WRAP-UP

Every lighting decision involves trade-offs:

For games/real-time:

• Lumen GI for realistic environments
• VSM shadows for broad hardware support
• Auto exposure often acceptable

For cinematics/product visualization:

• Direct lighting for control
• RT shadows for accuracy
• Manual exposure for consistency

For ML training data:

• Direct lighting only
• RT shadows/reflections
• Manual exposure — consistency is critical

1. Lighting is physics — Unreal simulates real light behavior
2. Reference materials — Known values (0.04, 0.18, 0.85) gauge your scene
3. Shadow softness = apparent source size — Bigger/closer lights = softer shadows
4. VSM vs RT — Different tools for different needs, check performance
5. Lumen is powerful but has trade-offs — Great for games, may not suit all workflows
6. Exposure control is critical — Auto vs Manual depends on your use case
7. Reflections reveal your lights — Understand this when designing lighting rigs

PBR values aren't made up — they come from scientific measurements of real materials. Here's how real-world properties become numbers in Unreal:

1. Spectrophotometer / Spectroradiometer

A spectrophotometer shines light at a material sample and measures what comes back:

• Illuminates sample with known light spectrum
• Measures reflected light at each wavelength
• Calculates percentage of light reflected = Reflectance

This gives you a number like "this material reflects 18% of incoming light."

2. Separating Diffuse from Specular

Here's the problem: when you photograph or measure a material, you capture both:

• Diffuse reflection — Light that enters the surface, scatters, and exits (color)
• Specular reflection — Light that bounces directly off the surface (shine)

For PBR, we need these separated.

Solution: Cross-Polarization

Studios like Activision (Call of Duty) and DICE use cross-polarized photography:

• Light source has polarizing filter (oriented one direction)
• Camera has polarizing filter (oriented 90° opposite)
• Specular reflections are polarized, so they get blocked
• Only diffuse light reaches the camera

This isolates the Base Color / Albedo — the diffuse component we need.

3. Reference Standards

To get accurate measurements, you need known reference points:

Calibrate your measurements against these, then measure your target materials.

4. Linear RGB Conversion

The measured reflectance percentage maps directly to Linear RGB:

Reflectance 18% → Linear RGB 0.18
Reflectance 85% → Linear RGB 0.85
Reflectance 4%  → Linear RGB 0.04

It's that simple. The percentage IS the linear value.

Our visual perception is non-linear. We're bad at judging actual reflectance:

• A surface that reflects 50% of light doesn't look "half as bright" as white
• Our brain compresses highlights and expands shadows
• We adapt to lighting conditions constantly

This is why artists historically made materials too dark or too bright — they painted what they perceived, not what the material actually does.

PBR fixes this by using measured values that respond correctly to any lighting.

If middle grey is "halfway between black and white," why is it 18% and not 50%?

Answer: Human perception is logarithmic, not linear.

But cameras are linear.

This mismatch is why gamma correction exists.

1. Light itself (Physics) — Linear

Double the photons = Double the energy

2. Camera sensors — Linear

Double the photons = Double the voltage = Double the pixel value

3. Human vision — Logarithmic

Double the photons = Slightly brighter (not twice as bright)

Camera sensors are just photon counters. They respond linearly:

• 100 photons → sensor value 100
• 200 photons → sensor value 200
• 1000 photons → sensor value 1000

If you displayed this raw linear data directly on a screen, it would look wrong:

• Shadows would be crushed to black
• Highlights would blow out
• Middle tones would look too dark

Why? Because our eyes expect a logarithmic response, not linear.

Gamma encoding (like sRGB) compensates for this mismatch:

Linear sensor data → Gamma curve → Looks correct to human eyes

The sRGB gamma curve (~2.2) redistributes the values:

• Stretches the dark tones (more bits for shadows)
• Compresses the bright tones (fewer bits for highlights)
• Makes 18% reflectance land at ~50% in the encoded file

This is why:

• Textures are stored as sRGB (gamma encoded) — optimized for human viewing
• Rendering happens in Linear — physically correct math
• Final output gets gamma corrected — back to what eyes expect

Unreal handles this automatically — it converts sRGB textures to Linear for rendering, then back to sRGB for display.

When we say "18% grey = 0.18 Linear RGB":

The 0.18 Linear value displays as roughly middle brightness because the display's gamma curve transforms it to match human perception.

Linear (How light actually works):

0% ────────────── 50% ────────────── 100%
Black            Half the photons        White

Perceptual (How we see it):

0% ─── 18% ─────────────────────────── 100%
Black   Looks like middle             White

Our eyes evolved to see detail across a huge range of light levels — from starlight to bright sun. To do this, our visual system compresses bright values and expands dark values.

Human brightness perception roughly follows a power curve (Stevens' Power Law):

Perceived Brightness ≈ (Actual Reflectance)^0.5

So:

• 18% reflectance → √0.18 ≈ 0.42 (42% perceived brightness)
• 50% reflectance → √0.50 ≈ 0.71 (71% perceived brightness — way brighter than "middle")

18% actual reflectance appears perceptually as middle grey.

In 1933, Munsell (the color system guy) found through experiments that a series of grey patches with reflectances of approximately 3%, 9%, 18%, 36%, 72%... appeared as evenly spaced brightness steps to human observers.

Notice: these aren't linear steps (10%, 20%, 30%...). They're roughly doubling each step — a logarithmic scale.

This is why camera light meters are calibrated to 18%:

• The meter assumes the scene averages to middle grey
• It sets exposure so that average scene = 18% reflectance output
• This puts "middle" in the middle of the camera's dynamic range

Problem: Point your camera at snow (90% reflectance), and the meter tries to make it 18% grey — underexposed. Point at coal (4% reflectance), and it overexposes to 18% grey.

This is exactly what we disable with Manual Exposure when consistency matters.

If your 18% grey sphere looks like "middle grey" on screen, your exposure is correct.

If it looks too dark → increase exposure compensation If it looks too bright → decrease exposure compensation

18% is not arbitrary. It's the reflectance value that human vision perceives as the midpoint between black and white. Every camera meter, every exposure system, and every PBR workflow is built around this fact.

These are measured real-world reflectance values for physically based rendering. In Unreal Engine's Metalness/Roughness workflow:

• Base Color = Albedo (diffuse reflectance for non-metals, reflection color for metals)
• Values are in Linear RGB (what Unreal uses internally)
• sRGB values shown for reference when working in Photoshop/image editors

Going outside these ranges creates materials that don't respond correctly to lighting.

For metals, Base Color represents reflection color, not diffuse.

For the crash course demonstration:

1. Non-metal Base Color should be 0.04 - 0.85 (Linear)
2. Metals should be 0.5 - 0.98 (Linear)
3. Nothing in nature is pure black (0) or pure white (1)
4. Middle grey is 0.18, not 0.5 (perception is non-linear)
5. Roughness doesn't change brightness (energy conservation)
6. If it looks wrong under different lighting, the values are wrong

• physicallybased.info — Searchable database with engine-specific values
• DONTNOD UE4 Chart — Industry-standard reference (Sébastien Lagarde)
• Substance PBR Guide — Comprehensive material creation guide
• Call of Duty: Advanced Warfare — Activision's measured material research