Why Lighting Math Still Belongs in an AI Video Workflow
Generative video has made it dramatically easier to produce a beautiful frame without owning a single light. It has not made it easier to produce a consistent frame. Ask anyone who has generated a twenty-shot sequence with a diffusion video model: shot three looks sunlit, shot seven looks like a basement, and shot twelve mysteriously changes the actor's skin tone. The frames are individually attractive and collectively unusable.
The fix is not a better model. It is a better mental model of light. When you understand how illuminance, aperture, shutter, and sensitivity relate to one another, you stop typing mood words and hoping. You start specifying a scene the way a director of photography does, and you get repeatability out of a stochastic tool.
The footcandle-to-f-stop relationship is the single most useful piece of that mental model. It is the bridge between a physical quantity (how much light lands on a subject) and a camera setting (how much of that light the lens lets through). This guide walks through the units, the math, the lookup tables, and the working habits that turn a meter reading into a prompt that survives twenty generations.
Footcandles and Lux: What the Numbers Actually Measure
The definition worth remembering
A footcandle is one lumen of luminous flux falling on one square foot of surface. That is it. It is an illuminance unit, meaning it describes light arriving at a plane, not light leaving a source and not light bouncing off a wall. A lux is the same idea in metric: one lumen per square meter.
Because the two units describe the same physical thing at different scales, conversion is pure arithmetic:
| Unit | Equivalent |
|---|---|
| 1 footcandle | 10.764 lux |
| 1 lux | 0.0929 footcandle |
| 10 footcandles | ~108 lux |
| 100 footcandles | ~1,076 lux |
| 1,000 footcandles | ~10,764 lux |
If you only memorize one number, memorize 10.76. Roughly, one footcandle is ten lux.
Real-world reference points
Numbers become useful only when you can anchor them. These are widely quoted illuminance values for common environments, and they are close enough for planning:
| Environment | Lux | Footcandles |
|---|---|---|
| Direct midday sun | ~100,000 | ~9,300 |
| Hazy or high sun | ~32,000 | ~3,000 |
| Bright overcast | 5,000-10,000 | 465-930 |
| Sunrise or sunset | ~400 | ~37 |
| Well-lit office | 320-500 | 30-46 |
| Living room, evening lamps | 100-150 | 9-14 |
| Twilight | ~10 | ~0.9 |
| Full moon | 0.1-0.3 | 0.01-0.03 |
Notice that the range spans six orders of magnitude. A candlelit interior and a desert at noon are not variations on a theme; they are different universes. Every exposure decision you make is a negotiation about which universe you are claiming to be in.
Incident versus reflected readings
A meter pointed at the light source measures incident illuminance, which is what footcandles naturally describe. A meter pointed at the subject measures reflected luminance, which depends on how bright the subject actually is. A white shirt and a black jacket under identical light produce wildly different reflected readings but identical incident readings.
This distinction matters enormously when you write prompts. When you describe a scene as 'a woman in a white shirt in a dim room', you have described two incompatible exposure problems. A generative model will resolve the contradiction arbitrarily. Deciding in advance whether you are exposing for the white shirt or letting it blow out is a directorial choice, and it is exactly the kind of choice that footcandle math makes explicit.
F-Stops: A Sequence, Not a Setting
The square-root rhythm
The f-number is the ratio of focal length to the diameter of the effective aperture. Because area scales with the square of diameter, halving the light requires dividing the f-number by the square root of two, roughly 1.41. That produces the familiar ladder:
f/1, f/1.4, f/2, f/2.8, f/4, f/5.6, f/8, f/11, f/16, f/22, f/32
Each step is exactly one stop, which means exactly half or double the light. The number is not a measurement of brightness; it is a position in a sequence. Thinking of f/2.8 as 'one stop faster than f/4' rather than 'a number smaller than four' is the difference between guessing and calculating.
T-stops and why cinema glass disagrees
Lenses absorb and reflect light internally, so a lens marked f/2.8 might transmit only as much light as an ideal f/3.2. Cinema lenses are therefore often marked in T-stops, which account for transmission. For planning purposes the distinction rarely changes your lighting plan by more than a third of a stop, but it explains why swapping lenses can shift your exposure even when every setting matches.
What an f-stop does beyond brightness
Aperture is a lighting control, but it is also a compositional control. Wide apertures shorten depth of field, isolating a face from a background. Narrow apertures lengthen it, keeping foreground and background legible together. When you choose f/2 instead of f/8, you are not just choosing an exposure; you are choosing how much of the world the audience is allowed to see sharply. That is a storytelling decision disguised as a technical one.
The Bridge: Connecting Illuminance to Aperture
The exposure equation in plain language
The relationship between the light on your subject and the aperture you need can be written compactly. For incident metering, using arithmetic ISO:
N² = (E × S × t) / C
Where N is the f-number, E is illuminance, S is the ISO value, t is shutter time in seconds, and C is a calibration constant that depends on whether you are working in lux or footcandles and on how your meter's receptor is shaped.
With the common convention of C = 250 for lux, the same constant becomes roughly 23 when you switch to footcandles:
N² ≈ (E_fc × S × t) / 23
That single line is the entire conversion table. Everything else is plugging in numbers.
A worked example
Suppose you have 100 footcandles on your subject, ISO 100, and a shutter of 1/48 second (the standard 180-degree shutter at 24 frames per second):
- N² = (100 × 100 × 0.0208) / 23
- N² = 208 / 23
- N² ≈ 9.05
- N = 3.0
So the correct aperture is f/2.8. Not f/4, not f/1.8. The math is unambiguous, and that is precisely why it is valuable: it removes an entire category of indecision from your shoot.
Why your meter may disagree
Calibration constants vary between manufacturers and between flat-disc and dome receptors, and ISO 2720 allows a range of legitimate values. Add in that lens markings are approximate, that ambient light drifts, and that a 'footcandle' measured on a gray card is not the same as one measured on a face, and you should expect real-world readings to land within about two-thirds of a stop of the formula.
The practical answer is not to chase precision. It is to build a table from your own meter, or to pick the formula value and then bracket a third of a stop in each direction. Either approach gets you a repeatable starting point, which is all a lighting plan ever needs.
Practical Lookup Tables You Can Actually Use
Baseline table: ISO 100, 1/48 second
| Footcandles | Approx. lux | EV at ISO 100 | Aperture |
|---|---|---|---|
| 5 | 54 | 4.4 | f/0.7 (impractical) |
| 10 | 108 | 5.4 | f/1.0 |
| 25 | 269 | 6.7 | f/1.4 |
| 50 | 538 | 7.7 | f/2 |
| 100 | 1,076 | 8.8 | f/2.8 |
| 200 | 2,153 | 9.8 | f/4 |
| 400 | 4,305 | 10.8 | f/5.6 |
| 800 | 8,611 | 11.8 | f/8 |
| 1,600 | 17,222 | 12.8 | f/11 |
| 3,200 | 34,444 | 13.8 | f/16 |
| 6,400 | 68,889 | 14.8 | f/22 |
The table has one elegant property: every doubling of footcandles is exactly one stop. Once you internalize that, you can extend it indefinitely in either direction without a calculator.
Shifting the table with ISO and shutter
Both ISO and shutter time move the answer in predictable stop-sized increments.
- Doubling ISO adds one stop of brightness, so the required aperture closes by a stop. Uncomfortably dim 100 fc scene at ISO 100 gives f/2.8; at ISO 400 it gives f/5.6.
- Halving shutter time removes one stop, closing the aperture by a stop. Switching from 1/48 to 1/96 turns f/2.8 into f/2.
- Doubling shutter time adds a stop, opening the aperture. Switching from 1/48 to 1/24 turns f/2.8 into f/4.
So the general rule is: count stops. Ten footcandles to one hundred footcandles is three and a bit stops of light, and you compensate with three and a bit stops of ISO, shutter, aperture, or neutral density.
Using neutral density deliberately
At 400 footcandles and ISO 100 you need f/5.6. If your creative intent demands f/1.8 for a shallow, dreamy portrait, you are roughly three stops too bright for that aperture. A 0.9 neutral density filter buys you exactly those three stops. In AI generation there is no filter, but there is a conceptual equivalent: you can specify 'shallow depth of field, warm practicals, subject slightly overexposed at the highlights' and get the same visual signature without pretending the physics vanished.
The Director's Role: Turning Mood into Numbers
Choosing a key level for each look
Every visual style has an implicit illuminance range. Directors and DPs rarely say 'three hundred footcandles'; they say 'soft, low-key, like late afternoon through net curtains', which happens to sit around 100 to 200 fc. Translating taste into ranges is the core directorial skill in lighting:
| Look | Typical key level | Aperture at ISO 100, 1/48 |
|---|---|---|
| Bright commercial, high key | 400-800 fc | f/5.6-f/8 |
| Natural daylight interior | 150-300 fc | f/2.8-f/4 |
| Moody drama, low key | 40-100 fc | f/2-f/2.8 |
| Film noir, hard shadows | 20-60 fc | f/1.4-f/2 |
| Candlelit, practical-driven | 2-10 fc | f/0.7-f/1 |
The candlelit row is instructive. It is genuinely hard to photograph candlelight at a normal frame rate without either a fast lens, a high ISO, or a lot of cheating. When a generative model produces a candlelit scene that looks clean and grain-free, it is quietly breaking the rules. That is fine for a stylized piece, but knowing it is happening lets you decide whether you want it.
Depth of field as a narrative tool
Once you know the aperture your light level implies, you inherit the depth of field that comes with it. This is where directors and DPs historically argued, and where AI workflows now quietly make the decision for you unless you intervene.
- A low-key scene at 40 fc forces a wide aperture, which gives you a soft background and a face-forward composition.
- A bright exterior at 3,000 fc forces a narrow aperture, which gives you deep focus and an environment that competes with the actor.
If your story needs deep focus in dim light, you have a genuine problem that requires more light, more sensitivity, or a different lens. Acknowledging that problem is what separates a plan from a wish.
From Meter Reading to Prompt: A Working Method
Step one: decide the world
Pick an illuminance band for the scene before you write a single word of prompt. 'Interior, late evening, practical lamps, roughly 10 to 15 footcandles' is a decision. Write it down where you can see it.
Step two: derive the camera settings
Use the table to find the aperture that the light level implies at your chosen ISO and shutter. Note the resulting depth of field. If it contradicts your story, adjust the light level, not the other settings.
Step three: translate into descriptive language
Generative models do not parse footcandles. They parse descriptions of light behavior. Translate the numbers into observable consequences:
- 1,000 fc with a hard source: 'crisp shadow edges, high contrast, bright highlights on skin, deep blue sky rendering, everything sharp from foreground to background'
- 100 fc with a soft source: 'gentle falloff, soft-edged shadows, warm skin tones, background softly out of focus, moderate contrast'
- 20 fc with a single hard practical: 'single dominant light source from frame left, long dramatic shadows, most of the frame falling into near-black, highlights pooling on cheekbones'
Notice that each description includes the consequence of the exposure rather than the exposure itself. That is what makes it useful to a model.
Step four: lock the language and repeat it
Consistency comes from repetition, not from cleverness. Build a short lighting block of twenty to forty words and reuse it verbatim across every shot in a sequence. Change only what genuinely changes: camera angle, subject position, action.
Case study: building a film noir look
The classic noir look is hard light, deep shadow, and a face half in darkness. In lighting terms that means a single small source, low overall illuminance, and an exposure that lets shadows crush.
A workable specification:
- Key level: 25-40 footcandles overall, with a hard key roughly four times brighter than fill
- ISO 100, 1/48 second, aperture around f/2 (25 fc) to f/2.8 (50 fc)
- Depth of field: shallow, background rendered as soft shapes
- Practical detail: visible shadows from window blinds, smoke or haze catching the beam
Translated into a reusable prompt block:
High-contrast noir lighting, single hard key light from the left, minimal fill, deep crushed shadows across half the face, visible blind shadows on the back wall, practical lamp glowing in the background, shallow depth of field, cool desaturated palette with warm practical accents.
That block specifies a physical situation, which is why it holds up across many generations. Now compare it with a vague equivalent such as 'dark moody cinematic lighting'. The vague version will drift: sometimes the model decides the scene is actually bright but color-graded dark, sometimes it fills the shadows, sometimes it invents window light. The precise version has fewer degrees of freedom for the model to wander into.
Common Mistakes and How to Catch Them
Confusing illuminance with luminance. Footcandles describe light hitting a surface. Brightness in the final image is a different quantity that depends on the surface. If a scene looks wrong, ask whether the problem is the light level or the reflectance of what is in the frame.
Treating ISO as free. Pushing ISO is a legitimate way to work in dim light, but it changes texture. If you specify a grainy, high-ISO look in one shot of a sequence and a clean, low-ISO look in the next, the sequence will feel stitched together. Pick an ISO band and stay inside it.
Forgetting the shutter. Shutter time is not just exposure; it is motion rendering. A 1/48 shutter at 24 frames per second gives natural motion blur. A 1/500 shutter gives a staccato, crisp look. Changing shutter to fix exposure silently changes how motion looks, and generative models respond to motion descriptions too.
Ignoring mixed light sources. Two sources of very different color temperature create a lighting problem that no aperture setting can solve. Decide whether you are balancing them or embracing the clash, and say so explicitly in the prompt.
Over-specifying numbers. No model needs to be told 'f/2.8 at ISO 100'. What it needs is the visual consequence: shallow focus, moderate contrast, soft shadow edges. Use the numbers to decide, then discard them when writing.
Assuming consistency comes from seed control. A fixed seed helps, but lighting language matters more. Two shots with a shared seed and different lighting descriptions will look like different films.
Frequently Asked Questions
Do I need a light meter to use any of this?
No. The lookup tables and the doubling rule cover most practical planning. A meter is useful if you are matching real footage to generated footage, but even then, a phone light meter app gets you within a usable range.
What is the single most useful shortcut?
Every doubling of footcandles equals one stop. Combine that with the ten-lux-to-one-footcandle approximation and you can estimate almost any exposure in your head.
How do I handle scenes where the light level changes during the shot?
Describe the change as a behavior, not a number: 'subject walks from a bright doorway into a dim hallway, exposure holding on the midtones, background falling two stops darker'. That gives a model a comprehensible transition.
Is a shallow depth of field always better?
No. It is a look, not an upgrade. Deep focus in bright light is a deliberate choice that lets environments carry meaning. The mistake is letting low light levels decide your depth of field by default.
What about HDR and log footage?
These affect how much latitude you have in post, not how much light reaches the sensor. They are useful concepts to reference when you want a flat, low-contrast image that implies grading will happen later.
Can I describe a specific lens instead?
Yes, and it is often more efficient. '35mm at f/1.4, close to the subject' communicates both field of view and depth of field in fewer words than an exposure explanation would.
Bringing It Together: A Light-First Habit
The footcandle-to-f-stop relationship is not nostalgia for the analog era. It is a compact description of how images get made, and compact descriptions are exactly what generative tools reward. When you know that a dim scene implies a wide aperture, that a wide aperture implies shallow focus, and that shallow focus implies a particular kind of composition, you can write prompts that are internally consistent instead of merely atmospheric.
Start with one habit: before generating anything, write down the illuminance band and the aperture it implies. Two lines. Then write your light description so that it produces the visual consequences of that combination, and reuse the description across every shot in the sequence. Your frames will stop looking like a mood board and start looking like a film.
