Add Aspect Ratio for 2step Latent Sample node and ICC profiles directory

nodes/aspect_ratio_2step_latent.py

@@ -0,0 +1,156 @@
+"""
+Aspect Ratio for 2step Latent Sample Node
+Provides width and height outputs for common aspect ratios, scaled to a base resolution.
+Optionally detects aspect ratio from an input image.
+"""
+
+import math
+from typing import Optional, Tuple
+
+class AspectRatio2StepLatent:
+    """
+    Calculate width and height for various aspect ratios at different base resolutions.
+    All outputs are divisible by 16 for optimal latent compatibility.
+    """
+    
+    # Base resolutions from 256 to 1024 in 64-step increments
+    BASE_RESOLUTIONS = [str(i) for i in range(256, 1025, 64)]
+    
+    # Aspect ratios from 1:2 to 2:1 with common values
+    # Format: "display_name": (width_ratio, height_ratio)
+    ASPECT_RATIOS = {
+        # Portrait ratios (height > width)
+        "1:2 Portrait": (1, 2),
+        "9:16 Portrait": (9, 16),
+        "2:3 Portrait": (2, 3),
+        "3:4 Portrait": (3, 4),
+        "4:5 Portrait": (4, 5),
+        # Square
+        "1:1 Square": (1, 1),
+        # Landscape ratios (width > height)
+        "5:4 Landscape": (5, 4),
+        "4:3 Landscape": (4, 3),
+        "3:2 Landscape": (3, 2),
+        "16:9 Landscape": (16, 9),
+        "2:1 Landscape": (2, 1),
+        # Additional common ratios
+        "9:21 Portrait (Ultrawide)": (9, 21),
+        "21:9 Landscape (Ultrawide)": (21, 9),
+        "1:1.414 Portrait (A4)": (1000, 1414),  # sqrt(2) approximation
+        "1.414:1 Landscape (A4)": (1414, 1000),
+    }
+    
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "base_resolution": (cls.BASE_RESOLUTIONS, {
+                    "default": "512",
+                    "tooltip": "Base resolution for the shorter side (256-1024 in 64-step increments)"
+                }),
+                "aspect_ratio": (list(cls.ASPECT_RATIOS.keys()), {
+                    "default": "1:1 Square",
+                    "tooltip": "Select aspect ratio. Ignored if image is connected."
+                }),
+            },
+            "optional": {
+                "image": ("IMAGE", {
+                    "tooltip": "Optional input image. If provided, its aspect ratio will be used instead of the dropdown selection."
+                }),
+            }
+        }
+    
+    RETURN_TYPES = ("INT", "INT", "STRING", "FLOAT")
+    RETURN_NAMES = ("width", "height", "ratio_string", "ratio_float")
+    FUNCTION = "calculate_dimensions"
+    CATEGORY = "image/preprocessing"
+    
+    DESCRIPTION = "Calculate width and height for 2-step latent sampling based on aspect ratio and base resolution. All outputs are divisible by 16."
+    
+    def _round_to_multiple(self, value: float, multiple: int = 16) -> int:
+        """Round a value to the nearest multiple of the specified number."""
+        return int(round(value / multiple) * multiple)
+    
+    def _get_ratio_from_image(self, image) -> Tuple[int, int]:
+        """Extract width and height from an image tensor and return as ratio."""
+        # Image tensor shape is typically [batch, height, width, channels]
+        if len(image.shape) == 4:
+            height = image.shape[1]
+            width = image.shape[2]
+        elif len(image.shape) == 3:
+            height = image.shape[0]
+            width = image.shape[1]
+        else:
+            raise ValueError(f"Unexpected image shape: {image.shape}")
+        
+        return width, height
+    
+    def _simplify_ratio(self, width: int, height: int) -> Tuple[int, int]:
+        """Simplify a ratio to its smallest form using GCD."""
+        gcd = math.gcd(width, height)
+        return width // gcd, height // gcd
+    
+    def _ratio_to_string(self, width_ratio: int, height_ratio: int) -> str:
+        """Convert ratio to a display string."""
+        simplified_w, simplified_h = self._simplify_ratio(width_ratio, height_ratio)
+        return f"{simplified_w}:{simplified_h}"
+    
+    def calculate_dimensions(self, base_resolution: str, aspect_ratio: str, image=None):
+        """
+        Calculate width and height based on aspect ratio and base resolution.
+        
+        Args:
+            base_resolution: Base resolution as string (256-1024)
+            aspect_ratio: Selected aspect ratio from dropdown
+            image: Optional input image tensor
+            
+        Returns:
+            Tuple of (width, height, ratio_string, ratio_float)
+        """
+        base_res = int(base_resolution)
+        
+        # Determine aspect ratio - from image if provided, otherwise from dropdown
+        if image is not None:
+            img_width, img_height = self._get_ratio_from_image(image)
+            width_ratio, height_ratio = img_width, img_height
+            ratio_source = "image"
+        else:
+            width_ratio, height_ratio = self.ASPECT_RATIOS[aspect_ratio]
+            ratio_source = "dropdown"
+        
+        # Calculate the ratio as a float (width/height)
+        ratio_float = width_ratio / height_ratio
+        
+        # Calculate dimensions using geometric mean approach
+        # This keeps the total pixel area roughly equal to base_res^2
+        # For a 4:5 ratio at 512 base: width = 512 * sqrt(4/5), height = 512 * sqrt(5/4)
+        # So width < 512 and height > 512 by proportional amounts
+        
+        aspect = width_ratio / height_ratio
+        # width = base_res * sqrt(aspect), height = base_res / sqrt(aspect)
+        # This ensures width * height ≈ base_res^2
+        sqrt_aspect = math.sqrt(aspect)
+        
+        width = self._round_to_multiple(base_res * sqrt_aspect)
+        height = self._round_to_multiple(base_res / sqrt_aspect)
+        
+        # Ensure minimum dimensions
+        width = max(width, 16)
+        height = max(height, 16)
+        
+        # Create ratio string
+        ratio_string = self._ratio_to_string(width_ratio, height_ratio)
+        if ratio_source == "image":
+            ratio_string = f"{ratio_string} (from image)"
+        
+        return (width, height, ratio_string, ratio_float)
+
+
+# Node registration
+NODE_CLASS_MAPPINGS = {
+    "AspectRatio2StepLatent": AspectRatio2StepLatent
+}
+
+NODE_DISPLAY_NAME_MAPPINGS = {
+    "AspectRatio2StepLatent": "Aspect Ratio for 2step Latent Sample"
+}