carloscape/octorato/image-transparency

Image Transparency Skill (Background Removal)

Removes solid backgrounds (like white paper from a scanned signature or white backgrounds from logos) and converts them to a transparent PNG.

When to use

• The user provides an image with a solid or grey/white background and needs it transparent.
• Adapting signatures or logos for documents (PDFs, watermarks) where the background interferes with the layout.

Workflow

1. Use Python with Pillow and NumPy to evaluate image threshold.
2. Mask pixels above the light threshold (background) and set their Alpha channel to 0.
3. Normalize ink/foreground pixels to a solid color if needed (e.g. ink blue #1a1a2e or pure black #000000).
4. Save as a new _transparent.png file or overwrite if requested.

Code Example (Python)

from PIL import Image
import numpy as np
def make_transparent(input_path, output_path, threshold=180, fg_color=None):
    img = Image.open(input_path).convert('RGBA')
    arr = np.array(img)
    
    # Convert to grayscale to find dark vs light
    gray = Image.open(input_path).convert('L')
    gray_arr = np.array(gray)
    
    # Background mask (light areas)
    bg_mask = gray_arr >= threshold
    arr[bg_mask, 3] = 0 # Set alpha to 0 completely transparent
    
    # Foreground mask (dark areas)
    if fg_color:
        ink_mask = gray_arr < threshold
        arr[ink_mask, 0] = fg_color[0]
        arr[ink_mask, 1] = fg_color[1]
        arr[ink_mask, 2] = fg_color[2]
        
    out_img = Image.fromarray(arr)
    out_img.save(output_path)
    print(f"Saved transparent image to {output_path}")
# make_transparent('firma.png', 'firma_trans.png', fg_color=(26, 26, 46))

Note on "nano banana": Historically triggered by dictation typos for "una nueva [skill]". This skill embodies that rule for simple scripted transparency!