About

PNG images store transparency in a fourth color channel - the alpha channel, where each pixel holds an opacity value from 0 (fully transparent) to 255 (fully opaque). Removing it incorrectly produces black artifacts where transparent pixels once existed, because most naive tools simply discard alpha without compositing against a background. This tool applies the standard alpha blending equation: C_out = C_src ⋅ α + C_bg ⋅ (1 − α), pixel by pixel across the entire image. All processing happens in your browser. No image data leaves your device.

The tool handles edge cases that spreadsheet-level approaches miss: premultiplied alpha artifacts, semi-transparent gradients, and anti-aliased edges. Output is available in PNG (lossless) or JPEG (lossy, configurable quality). Note: JPEG inherently has no alpha channel, so conversion to JPEG is itself a form of alpha removal - but without proper compositing the result contains black halos around formerly transparent edges. This tool prevents that.

Formulas

The alpha compositing operation uses the "over" operator from the Porter-Duff model. For each color channel of each pixel:

C_out = C_src × α255 + C_bg × (1 − α255)

Where C_src = source pixel channel value (R, G, or B), α = source alpha value (0 - 255), C_bg = background color channel value. The output alpha is forced to 255 (fully opaque). When α = 0, the pixel becomes pure background color. When α = 255, the pixel is unchanged. Intermediate values produce the smooth blended result that preserves anti-aliased edges.

For JPEG output, quality factor Q maps to the encoder's quantization table scaling. Higher Q values (0.85 - 0.95) preserve detail at larger file sizes. The relationship between Q and file size is non-linear - a quality of 0.80 typically produces files 40 - 60% smaller than 0.95.

Reference Data

Image Format	Supports Alpha	Channels	Typical Use Case	Max Colors	Compression
PNG-24	Yes (8-bit)	RGBA	Web graphics, logos	16.7M	Lossless (DEFLATE)
PNG-8	Yes (1-bit)	Indexed + Alpha	Simple icons	256	Lossless
JPEG	No	YCbCr (3)	Photos	16.7M	Lossy (DCT)
WebP (Lossy)	Yes (8-bit)	YUV + A	Web optimization	16.7M	Lossy (VP8)
WebP (Lossless)	Yes (8-bit)	RGBA	Web optimization	16.7M	Lossless
GIF	Yes (1-bit)	Indexed	Animations	256	Lossless (LZW)
BMP	Optional (8-bit)	RGBA or RGB	Legacy systems	16.7M	None
TIFF	Yes (8/16-bit)	RGBA	Print, archival	16.7M+	Various
SVG	Yes (via opacity)	Vector	Scalable graphics	Unlimited	None (XML)
ICO	Yes (8-bit)	RGBA	Favicons	16.7M	None/PNG
AVIF	Yes (8/10-bit)	RGBA	Next-gen web	16.7M+	Lossy/Lossless (AV1)
HEIF	Yes (8-bit)	RGBA	Apple ecosystem	16.7M	Lossy (HEVC)

Frequently Asked Questions

When alpha is discarded without compositing, transparent pixels retain their underlying RGB values - which are often (0, 0, 0) i.e. black. The alpha channel was masking those black pixels. Proper removal requires blending against a background color using the Porter-Duff "over" operator, which replaces masked areas with your chosen color instead of revealing the hidden black.

PNG output is lossless - the flattened image is pixel-perfect relative to the composited result. JPEG output introduces DCT compression artifacts, especially visible around sharp edges that were formerly anti-aliased against transparency. For logos and text, use PNG. For photographs with minor transparent regions, JPEG at quality 0.90 or above is acceptable.

Semi-transparent pixels (alpha between 1 and 254) are blended proportionally with the background color. A pixel with α = 128 on a white background gets 50% of its original color plus 50% white. This preserves smooth edge transitions instead of creating jagged stair-step artifacts that occur with binary alpha thresholding.

Browser canvas elements typically support up to 16384 × 16384 pixels, though the practical limit depends on available RAM. A 16384 × 16384 RGBA image requires approximately 1 GB of memory for its pixel buffer alone. The tool enforces a 16384px dimension limit and warns when images exceed 4096 × 4096 (approximately 67 MB buffer) to prevent browser tab crashes.

Yes. The operation is idempotent. If all pixels already have α = 255, the blending formula reduces to C_out = C_src × 1 + C_bg × 0 = C_src. The image passes through unchanged. This is useful when you need to guarantee a file has no alpha channel metadata - for instance, some print workflows reject PNGs that contain an alpha channel even if all pixels are opaque.

JPEG quality controls the quantization step applied to DCT coefficients. The relationship is non-linear: between quality 0.90 and 1.0, file size increases steeply because fewer coefficients are zeroed out. Between 0.50 and 0.70, size decreases are more gradual. The steepest size savings occur dropping from 0.95 to 0.85 - typically a 30-50% reduction with minimal perceptual loss.