Scheduled Shape Transform

Documentation

• Class name: SaltScheduledShapeTransformation
• Category: SALT/AudioViz/Scheduling/Image
• Output node: False

This node is designed for applying scheduled transformations to shapes over a sequence of frames. It enables dynamic adjustments of shape properties such as size, position, and rotation according to predefined schedules, facilitating complex animations and visual effects.

Input types

Required

• max_frames
  • Specifies the total number of frames for the shape transformation sequence, dictating the duration of the animation.
  • Comfy dtype: INT
  • Python dtype: int
• image_width
  • Defines the width of the output image frame.
  • Comfy dtype: INT
  • Python dtype: int
• image_height
  • Defines the height of the output image frame.
  • Comfy dtype: INT
  • Python dtype: int
• initial_width
  • Sets the initial width of the shape before transformation begins.
  • Comfy dtype: INT
  • Python dtype: int
• initial_height
  • Sets the initial height of the shape before transformation begins.
  • Comfy dtype: INT
  • Python dtype: int
• initial_x_coord
  • Determines the initial x-coordinate position of the shape.
  • Comfy dtype: INT
  • Python dtype: int
• initial_y_coord
  • Determines the initial y-coordinate position of the shape.
  • Comfy dtype: INT
  • Python dtype: int
• initial_rotation
  • Sets the initial rotation angle of the shape, in degrees.
  • Comfy dtype: FLOAT
  • Python dtype: float
• shape_mode
  • Specifies the mode of the shape to be transformed, affecting how the shape is interpreted and manipulated.
  • Comfy dtype: COMBO[STRING]
  • Python dtype: str

Optional

• shape
  • Optional. Provides the shape object to be transformed if not using a predefined shape mode.
  • Comfy dtype: MASK
  • Python dtype: object
• width_schedule
  • A schedule list defining the width transformation of the shape across the specified frames.
  • Comfy dtype: LIST
  • Python dtype: list[float]
• height_schedule
  • A schedule list defining the height transformation of the shape across the specified frames.
  • Comfy dtype: LIST
  • Python dtype: list[float]
• x_schedule
  • A schedule list defining the x-coordinate position transformation of the shape across the specified frames.
  • Comfy dtype: LIST
  • Python dtype: list[float]
• y_schedule
  • A schedule list defining the y-coordinate position transformation of the shape across the specified frames.
  • Comfy dtype: LIST
  • Python dtype: list[float]
• rotation_schedule
  • A schedule list defining the rotation angle transformation of the shape across the specified frames.
  • Comfy dtype: LIST
  • Python dtype: list[float]

Output types

• images
  • Comfy dtype: IMAGE
  • The output images resulting from the applied scheduled shape transformations, showcasing the dynamic changes in shape properties over the sequence of frames.
  • Python dtype: list[Image]

Usage tips

• Infra type: CPU
• Common nodes: unknown

Source code

class SaltScheduledShapeTransformation:
    @classmethod
    def INPUT_TYPES(cls):
        return {
            "required": {
                "max_frames": ("INT", {"min": 1}),
                "image_width": ("INT", {"default": 512, "min": -MAX_RESOLUTION, "max": MAX_RESOLUTION}),
                "image_height": ("INT", {"default": 512, "min": -MAX_RESOLUTION, "max": MAX_RESOLUTION}),
                "initial_width": ("INT", {"default": 256, "min": -MAX_RESOLUTION, "max": MAX_RESOLUTION}),
                "initial_height": ("INT", {"default": 256, "min": -MAX_RESOLUTION, "max": MAX_RESOLUTION}),
                "initial_x_coord": ("INT", {"default": 256, "min": -MAX_RESOLUTION, "max": MAX_RESOLUTION}),
                "initial_y_coord": ("INT", {"default": 256, "min": -MAX_RESOLUTION, "max": MAX_RESOLUTION}),
                "initial_rotation": ("FLOAT", {"min": 0, "max": 360, "step": 0.01}),
                "shape_mode": (["circle", "diamond", "triangle", "square", "hexagon", "octagon"], ),
            },
            "optional": {
                "shape": ("MASK", ),
                "width_schedule": ("LIST", ),
                "height_schedule": ("LIST", ),
                "x_schedule": ("LIST", ),
                "y_schedule": ("LIST", ),
                "rotation_schedule": ("LIST", ),
            }
        }

    RETURN_TYPES = ("IMAGE", )
    RETURN_NAMES = ("images", )
    FUNCTION = "transform_shape"
    CATEGORY = f"{MENU_NAME}/{SUB_MENU_NAME}/Scheduling/Image"

    def apply_multiply_operation(self, initial_value, schedule, frame_idx):
        factor = schedule[min(frame_idx, len(schedule) - 1)]
        return initial_value * factor

    def draw_shape(self, draw, shape_mode, center, width, height):
        if shape_mode == "circle":
            draw.ellipse([(center[0] - width / 2, center[1] - height / 2), (center[0] + width / 2, center[1] + height / 2)], fill="white")
        elif shape_mode == "square":
            draw.rectangle([(center[0] - width / 2, center[1] - height / 2), (center[0] + width / 2, center[1] + height / 2)], fill="white")
        elif shape_mode == "diamond":
            half_width = width / 2
            half_height = height / 2
            draw.polygon([center[0], center[1] - half_height, center[0] + half_width, center[1], center[0], center[1] + half_height, center[0] - half_width, center[1]], fill="white")
        elif shape_mode == "triangle":
            draw.polygon([(center[0], center[1] - height / 2), (center[0] + width / 2, center[1] + height / 2), (center[0] - width / 2, center[1] + height / 2)], fill="white")
        elif shape_mode == "hexagon":
            angle = 2 * np.pi / 6
            points = [(center[0] + math.cos(i * angle) * width / 2, center[1] + math.sin(i * angle) * height / 2) for i in range(6)]
            draw.polygon(points, fill="white")
        elif shape_mode == "octagon":
            angle = 2 * np.pi / 8
            points = [(center[0] + math.cos(i * angle) * width / 2, center[1] + math.sin(i * angle) * height / 2) for i in range(8)]
            draw.polygon(points, fill="white")

    def transform_shape(self, max_frames, image_width, image_height, initial_width, initial_height, initial_x_coord, initial_y_coord, initial_rotation, shape_mode, shape=None, width_schedule=[1.0], height_schedule=[1.0], x_schedule=[1.0], y_schedule=[1.0], rotation_schedule=[1.0]):
        frames = []
        for frame_idx in range(max_frames):
            width = self.apply_multiply_operation(initial_width, width_schedule, frame_idx)
            height = self.apply_multiply_operation(initial_height, height_schedule, frame_idx)
            x_coord = self.apply_multiply_operation(initial_x_coord, x_schedule, frame_idx)
            y_coord = self.apply_multiply_operation(initial_y_coord, y_schedule, frame_idx)
            rotation_fraction = rotation_schedule[min(frame_idx, len(rotation_schedule) - 1)]
            rotation_degree = rotation_fraction * 360

            img = Image.new('RGB', (image_width, image_height), 'black')
            if isinstance(shape, torch.Tensor):
                shape_image = mask2pil(shape)
                shape_image = shape_image.resize((max(int(width), 1), max(int(height), 1)), resample=Image.LANCZOS)
                rotated_shape_image = shape_image.rotate(rotation_degree, expand=True, fillcolor=(0), resample=Image.BILINEAR)
                paste_x = int(x_coord - rotated_shape_image.width / 2)
                paste_y = int(y_coord - rotated_shape_image.height / 2)
                img.paste(rotated_shape_image, (paste_x, paste_y), rotated_shape_image)
            else:
                shape_img = Image.new('RGBA', (max(int(width), 1), max(int(height), 1)), (0, 0, 0, 0))
                shape_draw = ImageDraw.Draw(shape_img)
                self.draw_shape(shape_draw, shape_mode, (shape_img.width / 2, shape_img.height / 2), width, height)
                rotated_shape_img = shape_img.rotate(rotation_degree, expand=True, fillcolor=(0), resample=Image.BILINEAR)
                paste_x = int(x_coord - rotated_shape_img.width / 2)
                paste_y = int(y_coord - rotated_shape_img.height / 2)
                img.paste(rotated_shape_img, (paste_x, paste_y), rotated_shape_img)

            frames.append(img)

        if frames:
            tensor = [pil2tensor(img) for img in frames]
            tensor = torch.cat(tensor, dim=0)
        else:
            raise ValueError("No frames were generated!")

        return (tensor, )