Skip to content

Blender 4.1: VFX & Video



  • A new Split node was added, which replaces the Split Viewer node with the exact same functionality, except it now provides its output as a node output. (153f14be2b)
  • A new Size input to the Kuwahara node was added to allow variable sizing. Additionally, a new High Precision option was added to the node in the Classic mode, which produces more accurate results for high resolution and high dynamic range images. (203559757a, e4a93d7b8c)
  • A new size property was added to the Pixelate node, which allows the node to be used without being surrounded by scale-down and scale-up nodes. (baab14ca38, feb2d02709)
  • A new Precision option was added to the node tree options. It allows choosing between full precision and automatic precision, where the automatic option uses half precision for the viewport and interactive compositing. Half precision runs faster with less memory usage, albeit with reduced precision and value ranges. (474b6fa070)
  • A new Filter Type option was added to the Map UV node to allow the user to choose between Anisotropic filtering and Nearest Neighbour filtering. (a3b7674c6e)


  • The Keying Screen node was changed to use a Gaussian Radial Basis Function Interpolation, which produces smoother temporally stable keying screens. The node now also have smoothness parameter to control the width of the interpolation. (75c947a467)

  • The Inpaint node was changed to use Euclidean distance instead of Manhattan distance, resulting in a more uniform filled regions. The node now also uses a two-pass algorithm that allows better smoothing for internal filling, as demonstrated in the following example. (48d7d60c96)

  • The Bilateral Blur and Bokeh Blur nodes now assume extended image boundaries, that is, they assume pixels outside of the image boundary are the same color as their closest boundary pixels. While they previously ignored such pixels and adjusted blur weights accordingly. (3d7e84f57d, a433adda8e)
  • The Double Edge Mask node now include the zero point of the gradient, while it previously started from the gradient value after zero. (049b0e6539)
  • The Flip node now works in local space, which means flipping will not change the location of the image if it was translated. (70a8a9e4d9)
  • The Crop node no longer flips its bounds if they were inverted, which means that if the upper bound was less than the lower bound, the node will crop the image to nothing. (4bf08198a7)
  • Scaling and Rotation in the Viewport Compositor are now immediately realized, which means scaling up an image will now actually produce more pixels, and that applying filters on rotated images will work as expected, instead of being applied along the direction of rotation. (e592763940)
  • Translations in the Viewport Compositor are now immediately realized for the axes that has enabled wrapping. Consequently, the image will not get translated, but its content will, in a clip on one side, wrap on the opposite side manner. Further, wrapping information is no longer propagated to future automatic realizations, so tilling or repeating an image is no longer possible. An alternative method of repetition will be introduced in a later patch. (1500a594ad)


  • The Viewport Compositor now cache multi-pass images, making multi-pass compositing much faster. (356480fabb)
  • The Double Edge Mask node is now orders of magnitude faster. (049b0e6539)
  • The Z Combine and Dilate nodes now use an improved Anti-Aliasing algorithm. (1aafb6802b)
  • The Defocus node now uses a more accurate method to compute the bokeh radius, so results should now match render engines better. (65e1a3a5c3)
  • The Sun Beams node now produces smoother results. (cbb738191e)
  • The compositor now only executes if its result is really used or viewed. (e165624885)


  • The Split Viewer node was removed, replaced by the Split node. (153f14be2b)



The video Sequencer got many performance optimizations across the board.


Image/movie filtering that is done when scaling/rotating strips has been improved.

  • Default strip filter is now "Auto", which automatically chooses the most appropriate filter based on scaling factors (PR#117853):
    • No scale/rotation and integer positions use Nearest,
    • Scaling up by more than 2x uses Cubic Mitchell,
    • Scaling down by more than 2x uses Box,
    • Otherwise Bilinear.
  • Strip transforms got Cubic filter option, previously it only existed in Transform Effect strip (PR#117100, PR#117517). Cubic filter exists in B-Spline (matches cubic elsewhere in Blender) and Mitchell (usually better for images) varieties. Cubic filtering is also faster now.
  • Bilinear filter no longer adds a transparent edge near image border when scaling it up. (PR#117717)
  • Subsampled3x3 filter was replaced by a more general Box filter, that better handles scaling images down by more than 3x. (PR#117584)
  • Various "off by one pixel" issues resulting in gaps between neighboring strips, or images being shifted by half a pixel, have been solved. (PR#116628)


Sequencer Scopes got visual look improvements (PR#116798):

Waveform (Luma)
Waveform (Parade)


  • Audio waveforms are now displayed by default, and got a display option to show upper half of the waveform only. (a95dd8438d) (1be8b51b11)