Lightweight ComfyUI node for text-prompted segmentation masks ◧ using CLIPSeg and mask auto-aligned trim. numpy 2.x safe, TensorFlow-free, auto model caching.

◧ AP ClipSEG Mask Light

Trims a mask region by percentage along the X and/or Y axis.

crop_x:
  Positive value → trim that percentage from the RIGHT edge.
  Negative value → trim from the LEFT edge.
  Example: 30 removes the rightmost 30 % of the mask.

crop_y:
  Positive value → trim from the BOTTOM.
  Negative value → trim from the TOP.
  Example: -40 removes the top 40 % of the mask.

auto_align:
  When True, trim axes rotate to match the mask's principal orientation
  (PCA major/minor axes of the masked region).  crop_y then always trims
  the 'bottom' of the masked shape, regardless of how it is rotated in
  the image — useful for tilted heads, diagonal limbs, etc.

feather:
  Soft-edge width in pixels at the cut boundary. 0 = hard cut.

mask_dilate:
  Expands the trimmed mask by this many pixels (morphological dilation).

mask_blur:
  Gaussian blur applied to the final mask after trimming and dilation.

ComfyUI Node: AP Mask Trim

Category

Inputs

Outputs

Extension: ◧ AP ClipSEG Mask Light

More nodes in ◧ AP ClipSEG Mask Light

Run ComfyUI workflows without the setup