Create Silhouette from Image

Strong contrast between subject and background is what matters most. Side profiles of people, animals, trees, and city skylines against a bright sky convert almost perfectly. Photos where the subject and background share similar brightness (a dark coat against a dark wall) force you to push the threshold slider to extremes and lose detail. If you control the shot, photograph the subject against a plain bright background and the conversion becomes one click.

Alpha mode uses each pixel’s alpha channel: alpha > threshold → silhouette. Only works on images with actual transparency (PNG/WebP with alpha channel). On a JPEG (all alpha = 255), every pixel becomes silhouette – useless. Luminance mode uses each pixel’s brightness: brightness < threshold → silhouette. Works on any image; produces a “dark-on-light = silhouette” effect like a stencil.

Rec 601 (default) is the classic NTSC formula. Rec 709 matches sRGB / modern displays. Average is simple R+G+B / 3. Max channel uses the brightest channel – useful when you want red text on a colored background to NOT become silhouette. Min channel uses the darkest channel – opposite use case.

In luminance mode: brightness 0-255. Threshold 128 (default) splits pixels into “dark half” (silhouette) and “light half” (background). Lower threshold makes only very dark pixels become silhouette; higher threshold makes more pixels become silhouette. In alpha mode: alpha 0-255. Threshold 128 means “more than half opaque”. Lower threshold = more pixels qualify; higher = fewer.

“X% silhouette” means X% of the image’s pixels qualified as silhouette. Very low (under 0.5%) or very high (over 99.5%) often means the threshold or mode is wrong for this image – the tool shows an info-toast warning in those cases.

No – silhouette is fundamentally lossy. Every pixel becomes one of two colors (silhouette or background). Detail inside the silhouette area is lost. That’s the point of a silhouette.

Format: {original-stem}-silhouette-{w}x{h}-{silHex}.{ext}. E.g., portrait-silhouette-1920x1080-000000.png. The hex makes it obvious which color you used.

10 MB. Larger files work in principle but Canvas getImageData gets memory-heavy.

No. Canvas API runs in the browser. Nothing leaves your device.