Pulling a performer out of their background, live, has historically meant one of two things: a green screen and the lighting budget that comes with it, or a depth camera and a second computer running software that speaks to your VJ rig over a network.
Resolume 7.28.0, released September 22, 2026, makes it an effect you drag onto a layer.
What shipped
In Wire, Resolume’s node-based patching environment, two new nodes:
- Human Segmentation
- Depth Estimation
In Arena and Avenue, effects and a blend mode built on those nodes:
- Background Removal
- Depth Estimation
- Depth Blur
- Depth Field
- Depth Displace
- Depth Pixelation
- Depth Shift
- Depth blend mode
Background Removal uses a human-segmentation model to separate subject from background, and — the part that matters in practice — lets you output the mask, invert it, smooth the edges and clean up the mask boundaries. A segmentation effect without mask controls is a demo; with them it’s a tool.
Depth Estimation generates a depth map from a 2D image, with temporal filtering and smoothing to reduce flickering between frames.
Background on Wire, the patching environment where the two new nodes live — not a demo of the 7.28 features themselves.
The temporal filtering is the unglamorous essential
Monocular depth estimation on video has one dominant failure mode: it’s computed per frame, so the depth map jitters. Small changes in lighting or noise shift the estimate, and anything driven by that map — a blur, a displacement — crawls and shimmers even when the subject is still.
For a still image nobody cares. For a live visual thrown across a stage at 60fps, it’s the difference between usable and unwatchable. Shipping temporal filtering and smoothing as part of the effect, rather than leaving VJs to build their own frame-blending workarounds, is the detail that says someone tested this in a room.
What it actually unlocks
The depth effects are more interesting than the segmentation, because a depth map is a control signal, not just a matte.
Once every pixel carries a distance value, you can drive parameters by depth: blur only what’s far away, displace geometry by proximity, pixelate the background while the performer stays sharp, composite two sources by relative depth with the new blend mode. That’s 2.5D compositing on a live camera feed, from software that was already running the show.
For installation and performance work, the practical win is one fewer machine. Depth-reactive visuals previously implied a Kinect or a RealSense, a second computer, a network protocol, and one more thing to fail during a show. Getting depth from an ordinary camera feed inside Resolume removes a whole tier of the rig.
The honest caveat: estimated depth is not measured depth. A monocular model infers distance from visual cues and will get confused by flat textures, unusual lighting and reflective surfaces in ways a time-of-flight sensor won’t. For expressive visual work that’s fine. For anything requiring the depth to be metrically correct, it isn’t a sensor replacement.