A photograph is flat, so a camera move across it just slides the whole picture. DEPTH EFFECTS estimates how far away every pixel is and then hands that estimate to effects that can use it — a parallax that parts near objects from far ones, and a LiDAR-style scan that crawls through the picture rather than across it.
It is its own inspector section rather than a pack inside Layer Effects, because everything in it has a prerequisite the other packs do not: the depth map. The section puts the Analyse depth button above its picker, so every effect inside it can assume the map exists.
Depth Effects work on still-image layers.
Analyse the photo
Select a photo layer, open DEPTH EFFECTS and click Analyse depth…. A depth model runs over the image and produces a depth map, which is stored inside the project.
- It runs once per photo. Afterwards the effects are free — nothing is re-analysed when you scrub, render or reopen the project.
- The first analysis of your session downloads the model (about 25 MB), so it needs an internet connection that one time. Your browser caches it, and the depth map itself is saved in the project, so renders and later sessions work offline.
- Analysis takes a second or two on a machine with a GPU, longer without one.
- The map travels with the project — duplicate the layer, copy it to another project, or export and reimport the board and the depth comes with it.
Use ↻ to re-analyse (after replacing the image, for example) and the reset to drop the map along with the effects using it — keeping an orphan map would be worse than dropping both.
Near Is decides which end of the scale is close: White is what the estimator produces, Black inverts it. Every Depth effect on the layer obeys the same switch, so they can never disagree about which end of the scene is near.
3D Depth
Gives the photo real parallax. The project camera drives it: the foreground travels further than the background, the way it would if you had filmed the scene.
Make it move
3D Depth does nothing on its own. Add a camera layer and give it a move:
| Camera move | Effect on a depth layer |
|---|---|
| Pan (Pan X / Pan Y) | Sideways parallax — the strongest, most obvious depth cue |
| Dolly | Pushes into the scene; near objects grow faster than far ones |
| Tilt | Look-around, as if leaning past the subject |
| Zoom | Nothing, by design — a zoom is a lens change, and a real lens zoom flattens a scene rather than revealing depth. Use dolly if you want to travel into the picture. |
A camera preset, a hand-keyframed camera or a Camera Motion Path all work. A channel with the camera lock on does not follow the camera, so its layers do not parallax either.
Controls
- Strength (default 100%, up to 300) — how far near and far pixels separate. Push past 150% for a stylised look; very high values on a busy photo start to smear the edges of foreground objects, which is the limit of what a single photo can be stretched to.
- Focus Plane (default 45%) — which depth stays pinned in place. Everything nearer than the focus plane moves against the camera, everything further moves with it. Lower it to make the background swim behind a locked subject; raise it to make a foreground object pop out over a locked background.
Both keyframe, so a rack focus on a photo is a keyframed Focus Plane. The effect also takes an Active Range, like an overlay effect — outside the window the strength is 0, as a hard edge, so keyframe Strength if you want a gentle arrival.
If you cannot see the parallax
The effect scales with the size of the camera move, because that is what it is reacting to. Check, in order:
- Is the camera actually moving? With the playhead inside a camera clip, Pan X / Pan Y / Dolly must change over the clip.
- Are you only zooming? Camera zoom is a lens zoom and deliberately produces no parallax. Swap it for Dolly, or add a pan.
- Is the move very small? The gentle handheld presets move Pan X by less than one percent — enough for a lifelike drift, but the depth separation at that scale is a few pixels.
- Is the channel camera-locked? A locked channel ignores the camera, so its layers cannot parallax.
- Is the subject where you think it is? Turn Strength to 300% temporarily — that makes the separation obvious and shows which parts of the picture the model judged near and far. Then dial it back.
A photo with clear foreground/background separation reads far better than a flat scene shot straight on. The model can only work with the depth cues that are in the picture.
Depth Scan
A LiDAR sweep through a photo rather than across it. Pixels whose depth is within the slice width of a moving threshold light up, so the contour crawls over the subject’s surface instead of marching across the frame — which is the whole difference between this and a wipe.
- Speed and Phase — the sweep is a pure function of clip time (
phase + speed × seconds, wrapped), so it is identical in a scrub, a range render and a full export. Sweep is Linear or Hold Ends, which makes it crawl at the extremes. - Slice Width — how thick a depth slice the scan lights.
- Spread — how far the colours spread through depth.
- Softness and Point Noise — the slice’s edge, and a scatter that reads as sampled points rather than a solid band.
- Colour — RGB Split (three slices a little apart in depth, which reads as chromatic fringing on the contour), Flat Color (one slice in a colour you pick) or Gradient (one wider slice with a ramp laid across it, far to near).
- Far Cut / Near Cut and Depth Curve — trim and reshape the depth range being swept.
- Border Reject — rejects the depth cliff at the subject’s silhouette. At 0 the silhouette becomes a permanent outline that never moves, which reads as a filter rather than a scan.
- Brightness, Original Image (how much of the photo shows under the scan) and Amount (0 is an exact bypass).
- View — Result, Depth Map or Original. The two diagnostics render and export exactly as shown, so set it back to Result before exporting.
Depth Scan takes an Active Range too, and inside a range the sweep clock is rebased so the slice starts at your Phase wherever the window sits rather than being caught mid-travel. Almost every number keyframes; Speed deliberately does not, because the phase is start + speed × time, so an animated rate would teleport the slice.
Depth Scan is free.
Where to go next
- Layer Effects
- Camera Layer — the camera that drives 3D Depth
- Motion Paths
- Keyframe Builder
Leave a Reply