DepthBlur

Resolve 21 places Depth Blur [DBl] in Deep Pixel Nodes. Here Deep means auxiliary per-pixel channels—most importantly Z-Depth—inside a normal 2D image stream. It is not a Deep Image node, so multi-sample deep data from DeepMerge/DeepToImage is not its direct input type.

DepthBlur has three physical inputs. The orange Input is the only required connection and supplies the image being blurred. The green Blur Image is an optional control image. The blue Effect Mask is an optional spatial gate. The old one-input article therefore missed one of the node's central workflows.

Blur Channel selects Z, Red, Green, Blue, Alpha or Luma. With Blur Image disconnected, the selected channel comes from the main Input. Once Blur Image is connected, the selected channel comes from Blur Image. A separate grayscale map can therefore drive the blur directly through Blur Image + Luma without first being packed into the beauty image.

Z is the focus-plane mode. Focal Point is visible only when Blur Channel is Z and selects the simulated focus distance. Lower values move the focus point closer to the camera and higher values move it farther away. Sample/Pick can read a Z-buffer value from the displayed image when valid Z data exists.

Depth of Field defines the depth of the region that stays focused around Focal Point. Blackmagic's example uses Focal Point 300 and Depth of Field 200, leaving Z values 200 through 400 in focus: the focal point sits at the middle of the range and DOF is its full width.

Z Scale rescales numeric Z-buffer values before depth response is evaluated. Raising it expands distances and lowering it contracts them. This is not Blur Size: Z Scale changes depth mapping while Blur Size changes horizontal/vertical blur strength. The Resolve 21 channel chapter specifically notes that floating-point images often need Z Scale far below the captured default 200 and gives about 0.2 as a useful experimental starting point.

Filter chooses Box, Soften or Super Soften. Box applies a basic depth-based box blur, Soften a general softening filter and Super Soften a higher-quality softening filter. Lock X/Y links horizontal and vertical Blur Size; unlock it only when asymmetric blur is actually intended.

Effect Mask is not the depth map. Resolve 21 states that an effect mask is applied after the tool is processed, so it only controls where the computed DepthBlur contributes. Depth/blur control remains a Blur Channel decision sourced from Input or Blur Image.

For a rendered Z workflow, Renderer3D/EXR must actually carry valid Z auxiliary data. When a separate Z pass needs to become an auxiliary Z channel on the beauty image, Channel Booleans or Copy Aux can build that structure. For general per-pixel blur, keeping a map separate and feeding Blur Image is a different valid topology.

A reliable Z setup order is: verify Z, use an obvious Blur Size, Sample Focal Point, calibrate Z Scale to the pass numeric range, then set Depth of Field and finally choose the finishing Filter/Blur Size. This keeps depth-mapping errors separate from simple blur-strength decisions.

The current Fusion 21 comparison lists 3D Z-Depth Blur among multi-channel capabilities for DaVinci Resolve Studio and Fusion Studio. This is a classic Fusion Deep Pixel workflow rather than a Studio-only Resolve FX category effect.

DepthBlur should not be confused with a normal Defocus node: here blur amount can vary per pixel from a selected channel, and Z mode adds a separate focus-plane model.

The term Deep Pixel refers to auxiliary channels carried by a normal image stream. This is distinct from Deep Image nodes where one pixel can hold multiple depth samples.

Main Input always remains the content image being blurred. Even when the control channel is read from the same input, its RGB is still the visual source of the result.

Blur Image is not a second beauty layer to blend. Once connected, it becomes the source of the selected Blur Channel rather than being added to the channel from Input.

For Red/Green/Blue/Alpha/Luma workflows, a separate Blur Image lets an artist-authored map drive blur without repacking the beauty image first.

With Blur Channel = Z, Focal Point, Depth of Field and Z Scale form a separate depth model. In RGB/A/Luma modes those Z-specific controls should not be treated as universal blur settings.

Focal Point is usually best acquired with Sample/Pick on the actual subject. Numeric Z units depend on the renderer/pass and rarely have a universal human-readable scale.

Depth of Field defines the range around Focal Point rather than maximum blur amount. Increasing DOF widens the sharp region instead of uniformly weakening blur.

Z Scale changes sensitivity to numeric distance values. A bad Scale can make almost the entire frame equally sharp or equally blurred even with a correct Focal Point.

Floating-point Z passes especially need calibration: the historical captured default of 200 can be orders of magnitude too high. A value around 0.2 is a manual diagnostic starting point, not a production preset.

Blur Size defines visible blur amplitude after channel/depth response has been established. It should not be used to compensate for incorrect Z Scale.

Lock X/Y helps separate depth response from directional styling. Keep it enabled during diagnosis and create asymmetric X/Y blur only after the depth model works.

Box, Soften and Super Soften change blur character but not the control-channel source. Changing Filter cannot repair missing Z or a wrong map.

Super Soften can cost more computation, so choose the final filter after channel/depth calibration rather than using high quality as a way to hide setup errors.

With a grayscale Blur Image and Blur Channel = Luma, map brightness drives blur. Red/Green/Blue modes can produce completely different patterns from the same map.

Alpha as Blur Channel is an image channel source. It is not the blue Effect Mask, which is a separate datatype and a separate processing stage.

Effect Mask acts as a spatial gate on the already processed result. It does not define Focal Point, DOF, Z Scale or the per-pixel blur map.

A soft Effect Mask can smoothly mix processed and unprocessed DepthBlur, but inside the admitted region the amount of blur still comes from the selected channel source.

For Renderer3D/EXR Z workflows, verify that Z actually exists and has the expected orientation. A correct beauty image says nothing about auxiliary Z availability.

If a separate Z pass must be packed into the beauty auxiliary channel, Channel Booleans/Copy Aux creates a different topology from using Blur Image directly. Those workflows should not be described as identical.

Blur Image + Luma is convenient for arbitrary artist maps; auxiliary Z + Z mode is appropriate when real Focal Point/DOF focus-plane semantics are needed.

At foreground/background depth discontinuities, post DOF can show halos because each pixel carries limited depth/coverage information. This is a post-process limitation rather than automatically a bad Z pass.

Hair, glass, motion blur and antialiased edges are especially difficult because coverage and depth do not always represent physically correct occlusion.

If Z direction from an external renderer is inverted or normalized differently, Focal Point/Scale can behave unexpectedly. Inspect the pass itself before blaming DepthBlur.

DepthBlur before Merge processes one plate/branch. After Merge it may read channels from an assembled composite where auxiliary Z has been lost or is no longer meaningful.

When 2D graphics are merged over 3D beauty, those graphics may have no valid Z. Applying DepthBlur after that merge requires deliberate auxiliary-channel management.

Common Settings Blend is a final mix and is not Blur Size. Blur Size changes the calculated blur, while Blend mixes the finished result with the source.

For an animated focus pull, animate Focal Point while keeping Z Scale/DOF/Blur Size stable unless the shot specifically requires those to change.

Rack-focus interpolation must be judged in motion because object Z values may be distributed nonlinearly across the scene.

Performance depends strongly on Filter, Blur Size, resolution and map complexity. Profile at target resolution before duplicating several DepthBlur nodes.

For automation, visible labels are insufficient: conditional controls, channel enums and internal IDs must be confirmed from a target-build .setting/.comp.

Static documentation defines topology and semantics, while exact edge sampling, conditional UI and render behavior remain part of live Resolve/Fusion certification.