Skip to content

Blender 4.4: Compositor

Added

  • Integer sockets are now supported in the compositor. (b3623feab2)
  • Exposed the Quality option of OpenImageDenoise on the Denoise Node. (8efd41271d)
    • This allows users to make trade offs between quality and performance of denoising in the compositor, similar to what can already be done for Cycles renders.

Changed

  • Transformations are now delayed until it is necessary to apply them. So scaling down an image then scaling it up again will no longer pixelate it, that is, transformations are no longer destructive until the image is actually processed. (109478d599)
  • The Wrapping option in the Translate node was turned into a Repeat option that infinitely repeats when mixed with a larger image. (5fd5685a6a)
  • The Glare node was revamped to provide a better user experience and more flexible control:
    • Node options are now single value inputs that can be linked. (004e3d39fa)
    • The generated glare and the highlights are now exposed as outputs. (004e3d39fa)
    • A new Strength input now controls the strength of the glare, and can boost the glare power. (004e3d39fa)
    • Fog Glow size is now linear, relative to the image size, and can have any size. (004e3d39fa)
    • Bloom size is now linear and relative to the image size. (004e3d39fa)
    • Bloom now conserves energy better and has a more reasonable output range. (b92a6eab3a)
    • The saturation and tint of the glare can now controlled using the new Saturation and Tint single value inputs. (c37d51f73d)
    • The highlights can now be clamped and smoothed using the new Smoothness and Maximum single value inputs. (3b28cf276e)
    • The inputs are now organized into panels for compactness and clarity. (14a380089a)

The new Glare node.

Compatibility

  • The Glare node options were turned into inputs, are now deprecated, and will be removed in a future version.
  • The Fog Glow and Bloom modes of the Glare node were corrected due to the revamped Glare node, but the input image size was assumed to be the render size during correction, since guessing the real input size was difficult. If this was not the case, turning the Size and Strength inputs should restore the old result.