Superside ComfyUI NodesandClaude Sonnet 5
5a721e11ea
New SupersideColorGradingNode: brightness/contrast/saturation multiplicative factors (via PIL ImageEnhance) plus additive per-channel R/G/B offsets, matching the common "Color Grading JK" behavior exactly (verified: zero pixel difference vs that node's logic) but vectorized with numpy instead of a per-pixel loop, so it's fast on large images. Purely local - no API key, no model call. Preserves alpha. Registered and documented. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
87 lines
3.6 KiB
Python
87 lines
3.6 KiB
Python
import logging
|
|
|
|
import numpy as np
|
|
import torch
|
|
from PIL import Image, ImageEnhance
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
|
class SupersideColorGradingNode:
|
|
"""
|
|
Color Grading Node: apply brightness, contrast, saturation and per-channel
|
|
R/G/B offset adjustments to an image. Purely local (no API, no key).
|
|
|
|
Behaviour matches the common "Color Grading" utility:
|
|
- brightness / contrast / saturation are multiplicative factors
|
|
(1.0 = no change) applied via PIL's ImageEnhance.
|
|
- R / G / B are additive offsets (-255..255) added to each colour
|
|
channel and clamped to 0-255.
|
|
Alpha is preserved if the input has it.
|
|
"""
|
|
|
|
CATEGORY = "Superside"
|
|
|
|
@classmethod
|
|
def INPUT_TYPES(cls):
|
|
return {
|
|
"required": {
|
|
"image": ("IMAGE",),
|
|
"brightness": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 3.0, "step": 0.01,
|
|
"tooltip": "Brightness factor. 1.0 = no change, <1 darker, >1 brighter."}),
|
|
"contrast": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 3.0, "step": 0.01,
|
|
"tooltip": "Contrast factor. 1.0 = no change."}),
|
|
"saturation": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 3.0, "step": 0.01,
|
|
"tooltip": "Saturation factor. 1.0 = no change, 0 = grayscale."}),
|
|
"R": ("INT", {"default": 0, "min": -255, "max": 255, "step": 1,
|
|
"tooltip": "Red channel offset added to every pixel (clamped 0-255)."}),
|
|
"G": ("INT", {"default": 0, "min": -255, "max": 255, "step": 1,
|
|
"tooltip": "Green channel offset added to every pixel (clamped 0-255)."}),
|
|
"B": ("INT", {"default": 0, "min": -255, "max": 255, "step": 1,
|
|
"tooltip": "Blue channel offset added to every pixel (clamped 0-255)."}),
|
|
},
|
|
}
|
|
|
|
RETURN_TYPES = ("IMAGE",)
|
|
RETURN_NAMES = ("image",)
|
|
FUNCTION = "color_grade"
|
|
DESCRIPTION = (
|
|
"Apply brightness, contrast, saturation and per-channel R/G/B offsets "
|
|
"to an image. Local color grading - no API key, no model call."
|
|
)
|
|
|
|
def color_grade(self, image, brightness, contrast, saturation, R, G, B):
|
|
out_frames = []
|
|
|
|
# A ComfyUI IMAGE batch is (B, H, W, C) float 0-1.
|
|
for frame in image:
|
|
frame_np = frame.cpu().numpy() if isinstance(frame, torch.Tensor) else np.asarray(frame)
|
|
has_alpha = frame_np.ndim == 3 and frame_np.shape[2] == 4
|
|
|
|
rgb_u8 = (np.clip(frame_np[..., :3], 0.0, 1.0) * 255.0).astype(np.uint8)
|
|
pil = Image.fromarray(rgb_u8, mode="RGB")
|
|
|
|
if brightness != 1.0:
|
|
pil = ImageEnhance.Brightness(pil).enhance(brightness)
|
|
if contrast != 1.0:
|
|
pil = ImageEnhance.Contrast(pil).enhance(contrast)
|
|
if saturation != 1.0:
|
|
pil = ImageEnhance.Color(pil).enhance(saturation)
|
|
|
|
arr = np.array(pil).astype(np.int16)
|
|
if R != 0:
|
|
arr[..., 0] = np.clip(arr[..., 0] + R, 0, 255)
|
|
if G != 0:
|
|
arr[..., 1] = np.clip(arr[..., 1] + G, 0, 255)
|
|
if B != 0:
|
|
arr[..., 2] = np.clip(arr[..., 2] + B, 0, 255)
|
|
arr = arr.astype(np.float32) / 255.0
|
|
|
|
if has_alpha:
|
|
alpha = np.clip(frame_np[..., 3:4], 0.0, 1.0).astype(np.float32)
|
|
arr = np.concatenate([arr, alpha], axis=-1)
|
|
|
|
out_frames.append(torch.from_numpy(arr).unsqueeze(0))
|
|
|
|
return (torch.cat(out_frames, dim=0),)
|