What Is a Shallow Depth of Field?

What it buys and what it costs

Isolation is the point. A face against a busy street is competing with the street until the street stops being legible, and defocus is the only in-camera tool that removes background information without removing the background. It also reads as production value, because a thin sharp zone requires fast glass, controlled distance, and someone paying attention to focus.

The cost is information. Everything you blurred is context you can no longer use: the location, the time of day, the other characters, the joke in the background. Comedies and ensemble scenes routinely refuse the look for exactly this reason. It also flattens depth cues, so a shot with a thin sharp zone can feel like a subject floating in front of a screen rather than standing in a place.

There is a second cost that only shows up on set: margin. A thin sharp zone means any small error, an actor shifting their weight, a slight recompose, a breath, becomes a soft shot.

Building it

Four inputs stack, and it is easier to think of them as a budget than as one dial.

• Open the aperture. The most direct move, and the one with a hard limit set by whatever glass you have.
• Use a longer lens and step back. An 85mm from four metres separates far better than a 35mm from one metre, and it also flatters faces.
• Get closer. The strongest lever per unit of effort, and free.
• Push the background away. Frequently forgotten and often the biggest win. A subject standing a metre from a wall cannot have a blurred background at any aperture. Move them six metres out and the same lens does the work for you.

If the goal is separation rather than blur specifically, the fourth item deserves to be tried first, because it costs nothing and does not eat your focus margin.

How generated images handle it

Models overshoot in this direction by default. Training captions that mention portrait, cinematic, or professional are overwhelmingly attached to images with destroyed backgrounds, so the model has learned that blur means good. The result is that you usually have to ask for less, not more.

When you do want it, describe the boundary rather than naming the effect:

• ✅ 85mm at f/1.4, both eyes sharp, background reduced to soft shapes
• ✅ sharp on the hands in the foreground, face already softening behind them
• ⚠️ shallow depth of field alone gives an unpredictable amount of blur, often too much
• ⚠️ blurry risks motion blur or a soft subject instead

The reason naming the sharp part works better is that models render content confidently and infer depth poorly. Telling it what must stay crisp gives it an anchor. Telling it to blur gives it permission to blur anything.

Checks worth doing

Zoom in on the transition. The reliable failures are a sharp halo tracing the subject outline, hair or fingers assigned to the wrong plane, and a background that is uniformly soft in a way no real lens produces, with no gradient from near to far. Also check text: a sign well behind the subject should be unreadable, and if it is crisp the blur was painted rather than derived.

For video, watch a clip through twice. The blur boundary often crawls, and background elements can shift between sharp and soft with no camera movement to justify it. Slower moves and shorter durations both reduce it, and a moderately less extreme look is far more likely to survive a full clip intact.