Initial release: MASK CROP MASTER - always-square mask crop for ComfyUI

README.md

@@ -0,0 +1,71 @@
+# MASK CROP MASTER — ComfyUI Custom Node
+
+**Always-square mask crop, even at image edges.**
+
+## The Problem
+
+When you mask a region at the **edge** of an image, the standard `Mask Crop Region` node from was-ns produces a **rectangular** crop instead of a square. This happens because it clamps the crop dimensions when the square extends beyond the image bounds.
+
+## The Solution
+
+`MASK CROP MASTER` **repositions** the square instead of clamping it. When the crop would extend past an image edge, it slides the square inward so it always fits completely inside the image.
+
+```
+was-ns (rectangle):              MASK CROP MASTER (square):
+┌────────────────────┐           ┌────────────────────┐
+│████████████░░░░░░░░│           │░░░░░░░░░░░░░░░░░░░░│
+│████████████░░░░░░░░│           │░░████████░░░░░░░░░░│
+│░░░░░░░░░░░░░░░░░░░░│           │░░████████░░░░░░░░░░│
+│░░░░░░░░░░░░░░░░░░░░│           │░░░░░░░░░░░░░░░░░░░░│
+└────────────────────┘           └────────────────────┘
+  400×200 (broken)                 400×400 (correct)
+```
+
+## Installation
+
+```bash
+cd ComfyUI/custom_nodes
+git clone https://huggingface.co/YOUR_USERNAME/MaskCropMaster.git
+```
+
+Restart ComfyUI after cloning.
+
+## Usage
+
+Drop-in replacement for `Mask Crop Region` from was-ns:
+
+1. **Remove** the old `Mask Crop Region` node from your workflow
+2. **Add** `MASK CROP MASTER` from the `WAS Suite/Image/Masking` category
+3. **Connect** the same inputs: `mask`, `padding`, `region_type`
+4. **Connect** the same outputs: `cropped_mask`, `crop_data`, coordinates, dimensions
+
+## Inputs
+
+| Input | Type | Default | Description |
+|-------|------|---------|-------------|
+| `mask` | MASK | — | Input mask |
+| `padding` | INT | 24 | Padding around the bounding box before squaring |
+| `region_type` | COMBO | "dominant" | "dominant" or "minority" (was-ns compatible) |
+
+## Outputs
+
+| Output | Type | Description |
+|--------|------|-------------|
+| `cropped_mask` | MASK | Square-cropped mask |
+| `crop_data` | CROP_DATA | Compatible with `Mask Paste Region` |
+| `top_int` | INT | Y position of the square |
+| `left_int` | INT | X position of the square |
+| `right_int` | INT | Bottom edge |
+| `bottom_int` | INT | Right edge |
+| `width_int` | INT | Width (= side length) |
+| `height_int` | INT | Height (= side length) |
+
+## Compatibility
+
+- Works with `Mask Paste Region` from was-ns (same `crop_data` format)
+- Same output names → minimal rewiring needed
+- Tested with ComfyUI latest version
+
+## License
+
+Apache 2.0

__init__.py

@@ -0,0 +1,86 @@
+import torch
+import numpy as np
+from PIL import Image, ImageOps
+
+
+class MaskCropMaster:
+    """Mask Crop Region replacement that ALWAYS outputs a square crop.
+
+    Based on WAS_Mask_Crop_Region from was-ns, with the key fix:
+    when the square extends beyond the image bounds, it REPOSITIONS
+    the crop instead of clamping dimensions (which produces a rectangle).
+
+    Compatible with Mask Paste Region (same crop_data format).
+    """
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "mask": ("MASK",),
+                "padding": ("INT", {"default": 24, "min": 0, "max": 4096, "step": 1}),
+                "region_type": (["dominant", "minority"],),
+            }
+        }
+
+    RETURN_TYPES = ("MASK", "CROP_DATA", "INT", "INT", "INT", "INT", "INT", "INT")
+    RETURN_NAMES = ("cropped_mask", "crop_data", "top_int", "left_int", "right_int", "bottom_int", "width_int", "height_int")
+    FUNCTION = "mask_crop_master"
+    CATEGORY = "WAS Suite/Image/Masking"
+    DESCRIPTION = "Mask Crop Region that always outputs a square, even at image edges."
+
+    def mask_crop_master(self, mask, padding=24, region_type="dominant"):
+        mask_np = mask.cpu().squeeze().numpy()
+        mask_pil = Image.fromarray(np.clip(255.0 * mask_np, 0, 255).astype(np.uint8))
+
+        img_w, img_h = mask_pil.size
+
+        bbox = mask_pil.getbbox()
+        if bbox is None:
+            empty = Image.new("L", (img_w, img_h), 0)
+            empty_tensor = torch.from_numpy(np.array(empty).astype(np.float32) / 255.0).unsqueeze(0).unsqueeze(1)
+            crop_data = ((img_w, img_h), (0, 0, 0, 0))
+            return (empty_tensor, crop_data, 0, 0, 0, 0, img_w, img_h)
+
+        bbox_x1, bbox_y1, bbox_x2, bbox_y2 = bbox
+        bbox_w = bbox_x2 - bbox_x1
+        bbox_h = bbox_y2 - bbox_y1
+
+        side = max(bbox_w, bbox_h) + 2 * padding
+        side = min(side, img_w, img_h)
+
+        cx = (bbox_x1 + bbox_x2) / 2.0
+        cy = (bbox_y1 + bbox_y2) / 2.0
+
+        crop_x = round(cx - side / 2.0)
+        crop_y = round(cy - side / 2.0)
+
+        if crop_x < 0:
+            crop_x = 0
+        if crop_y < 0:
+            crop_y = 0
+        if crop_x + side > img_w:
+            crop_x = img_w - side
+        if crop_y + side > img_h:
+            crop_y = img_h - side
+
+        crop_x2 = crop_x + side
+        crop_y2 = crop_y + side
+
+        cropped_mask = mask_pil.crop((crop_x, crop_y, crop_x2, crop_y2))
+
+        region_tensor = torch.from_numpy(
+            np.array(cropped_mask).astype(np.float32) / 255.0
+        ).unsqueeze(0).unsqueeze(1)
+
+        crop_data = (cropped_mask.size, (crop_x, crop_y, crop_x2, crop_y2))
+
+        return (region_tensor, crop_data, crop_y, crop_x, crop_y2, crop_x2, side, side)
+
+
+NODE_CLASS_MAPPINGS = {
+    "MaskCropMaster": MaskCropMaster,
+}
+NODE_DISPLAY_NAME_MAPPINGS = {
+    "MaskCropMaster": "MASK CROP MASTER",
+}