Contrast is the actual job
A common misreading is that fill light exists to make the image brighter. It does not. Overall brightness belongs to exposure and to the key. What a fill light controls is the distance between the lit side and the shadow side, and that distance is what the audience reads as mood.
The measurement is a ratio, and it is easier in stops than in numbers. One stop of difference is 2:1, two stops is 4:1, three stops is 8:1. A face at 2:1 looks pleasant and informational. The same face at 8:1 looks like it is keeping something back. Nothing moved except one dimmer.
Because it is a ratio, you can get there from either direction. Turning the fill down and turning the key up both increase contrast. On location, where the key is often a window you cannot dim, adjusting the fill is the only lever you have.
Placement rules
- Near the lens axis, opposite the key. From that position its shadows hide behind the subject. The further round it goes, the more it behaves like a second key, which is where crossed nose shadows come from.
- Big and soft. A hard fill draws attention to itself. Bounce it off a card, a wall or a ceiling and the shadow edges it creates become too soft to notice.
- Dimmer, by distance if possible. Pulling a source back cuts its output and softens its shadow edge at the same time, which is usually what you want.
- Watch the catchlights. Two bright reflections in each eye is the signature of a visible fill, and it makes a portrait look lit rather than observed. Lower it, soften it, or accept only the key registering in the eyes.
Colour matters more than people expect. A fill that is slightly cooler than the key reads as natural, because in the real world the sky fills what the sun keys. A fill warmer than the key tends to look muddy on skin, since the shadow side loses its blue and the whole face collapses toward one hue.
When to use none at all
Plenty of good work has no fill in it. Interiors with light walls fill themselves. Exteriors have the sky, the ground and every white surface around, which is why on an overcast day the problem is usually too much fill rather than too little. That is when you reach for negative fill: black cloth on the shadow side, close enough to matter, absorbing what the environment is throwing back.
The habit worth building is to look at the shadow side first and decide what it should be, then choose the tool. Adding light is one option. Taking it away is the other, and it is the one that separates a shot that looks lit from one that looks like a room.
Prompting fill in AI generation
Models are biased toward flat light, because most of the images they learned from were flat. Ask for a portrait with no lighting instruction and you get an evenly filled face with no shadow side at all. So the useful instructions here are mostly subtractive.
- ✅
no fill light, shadow side of the face falling to near black - ✅
soft key from a window at frame left, gentle bounce fill on the right one stop down, shadow detail retained - ⚠️
fill lighton its own usually just brightens the whole frame and cancels the direction you asked for - ⚠️
evenly lit, no harsh shadowsproduces exactly the flat default you are trying to escape
Ratios in words work better than ratios in numbers. Models do not meter, so one stop down is a weaker instruction than shadow side visibly darker but still holding detail. If you want the shadow side to keep information, say what should stay visible in it, for example the eye on the dark side or the texture of a jacket. And in video, a shadow side tends to lift over the length of a clip as the model averages toward its default, so it is worth stating that the contrast holds to the end.