Channel Buttons / Red, Green, Blue, Alpha
Select which channels receive the custom filter.
← Fusion node reference
Image ProcessingAdvanced
CustomFilter
Programmable convolution kernel
Applies a custom convolution matrix for blur, sharpen, emboss, edge and other neighborhood filters.
CustomFilter
DaVinci Resolve · Fusion
CustomFilter exposes the convolution kernel directly, allowing you to design your own pixel-neighborhood operation rather than choosing a fixed filter preset.
Overview
What CustomFilter does
Each output pixel is calculated from nearby source pixels multiplied by the values in the matrix. Changing the kernel can create blur, sharpening, embossing, directional edge detection and many specialized technical filters.
Normalize is useful when the matrix changes total image energy. Without normalization a kernel can brighten or darken the image unexpectedly.
Because custom kernels can amplify noise and edge ringing, the safest workflow is to begin with a small matrix and validate the result on high-contrast details.
Mental model
Think of the matrix as a stencil of weights placed over every pixel: the stencil decides how much each neighbor contributes to the output.
Common uses
Custom sharpening
Emboss and relief effects
Directional edge detection
Special blur kernels
Technical image-processing experiments
Flow anatomy
Inputs & outputs
The port structure shows how this node fits into a Fusion graph.
Image
The source image to process with the convolution matrix.
CustomFilter
Fusion node
Output
The processed image after the node operation.
Inspector
Important parameters
Start with these controls before reaching for more complicated workarounds.
Matrix Size
Sets the dimensions of the convolution kernel.
Matrix Values
Defines the weight of each neighboring sample.
Update Lock
Prevents constant recalculation while editing a complex matrix.
Normalize
Normalizes the kernel so overall brightness remains more stable.
Floor Level
Sets a lower output floor for the filtered result.
Practical setup
A reliable workflow
- 1
Choose the channels that should be processed.
- 2
Start with a small Matrix Size.
- 3
Enter the kernel values and enable Normalize when appropriate.
- 4
Inspect fine edges and noise at 100 percent.
- 5
Mask or blend the result if the custom kernel is too aggressive.
Typical node chain
Neighborhood Pixels→
Custom Matrix→
Filtered Image
Examples
Where this node fits in real work
Custom edge kernel
Build an edge-emphasis matrix for a technical matte or stylized look.
MediaInCustomFilterBitmap/GradeMediaOut
Emboss branch
Use an asymmetric kernel to create directional relief.
MediaInCustomFilterMergeMediaOut
Troubleshooting
Common mistakes
Using an unnormalized kernel and mistaking brightness shifts for the intended effect.
Applying large kernels to noisy material without checking ringing.
Filtering alpha unintentionally.
Production notes
Practical tips
Keep useful kernels documented so they can be reused.
Test channels independently before processing RGBA together.
Combine with a mask for local detail enhancement.
Related nodes