Depth of Field
A per-pixel circle of confusion computed from the depth inferred out of your flat photo, gathered through a configurable aperture — polygonal, catadioptric or anamorphic bokeh.
← All the FXIn the app this is called Depth DOF.
- Family
- DEPTH ✦
- Availability
- Part of the full library — see pricing
Why most background blur doesn’t look like a lens
The usual approach is a radial gradient: sharp in the middle of the frame, increasingly blurred toward the edges. It is quick and it is wrong, and everyone can tell without being able to say why.
A lens is not sharp in the MIDDLE. A lens is sharp at a DISTANCE. Something at the very edge of the frame that happens to sit at the focus distance is perfectly sharp, and something dead centre that happens to be far away is not. Get that backwards and every image betrays itself the moment the subject isn’t centred, or the moment anything at the subject’s distance appears near the frame edge.
Doing it properly needs a depth value per pixel, which is exactly what this family has: the depth is inferred from the image itself, once, in your browser, and cached for every depth effect to share.
Circle of confusion, and why bokeh has a shape
A point of light that isn’t in focus does not become a soft smudge. It images as a disc — and the disc is the shape of the aperture the light came through. Its diameter grows with how far the point sits from the focal plane, in either direction, which is why foreground and background both go soft while the plane between them stays sharp. That disc is the circle of confusion, and computing it per pixel from the depth is what makes the falloff behave.
The shape is then the aperture’s business. An iris built from straight blades renders out-of-focus highlights as polygons, and the blade count decides whether they read as hexagons, heptagons or near-circles — the same geometry that turns bright points into sunstars when the lens is stopped down. A catadioptric — a mirror lens — has a secondary mirror blocking the centre of its aperture, so its out-of-focus highlights are rings rather than discs; that donut bokeh is a structural consequence of the design, not a defect. An anamorphic squeezes one axis, so its discs come out as ovals.
Weighting the gather toward highlights matters for the same reason bokeh is only ever discussed in terms of highlights: a dim point spread into a disc contributes almost nothing to what it lands on, and only a bright one survives being spread thinly enough to be seen as a shape.
What the inferred depth can and can’t do
The depth channel is inferred, not measured. It is very good at the thing that matters here — ordering the scene front to back — and it is less reliable at fine, thin structures like hair, wire fencing and foliage edges, where a real depth sensor would also struggle.
In practice that means the focal-plane and falloff controls do more for believability than the aperture shape does. Set the plane where the subject actually is and let the transition be gradual, and the inference’s soft edges hide inside a soft transition. Ask for a hard, shallow separation on a subject with complicated edges and you are asking the depth to be more precise than it is.
If you want to see what the effect is working from, Z-Depth renders the depth channel directly — false colour, contours or displacement — which is the fastest way to understand why a particular image is or isn’t separating.
When you’d reach for it
When a flat photograph needs to feel like it was taken through glass, and when you need the separation to hold as the subject moves around the frame. It is also the honest way to draw attention: blur is the only tool that removes information rather than adding it, which is why it directs the eye so reliably.
Related effects
Open the app to try it; the free effects work with no signup. Open V3KTR · all the FX · how it works