Image guide · Updated September 19, 2026
Keep transparent edges while reducing PNG size
A transparent logo and a detailed photograph have very different compression needs. Preserve the parts that make the image useful.
Lossless describes the optimization step
MallowDesk uses lossless PNG optimization after decoding and any resize you explicitly request. Original-size PNG optimization preserves the prepared pixel grid. Resizing, orientation correction, and color conversion happen before that step; lossless does not mean every source byte or metadata field is retained. Output is an 8-bit sRGB workflow.
Why a PNG can stay large
Detailed textures and noisy pixels are hard to encode compactly without losing information. A quality slider cannot guarantee that lossless PNG reaches 100 KB. Target mode tries a PNG encode and reports a failure if the final file is too large. Reducing dimensions removes pixels. Converting to WebP or AVIF may retain alpha while changing pixel values through lossy compression.
JPEG has no alpha channel
When converting a transparent image to JPEG, choose Fill transparency with a background color and select the actual background used by your page or document. The tool stops and explains the problem if that choice is missing. Once flattened, a logo with a white background will show a white rectangle on a dark page; changing the viewer background will not restore transparency.
Check edges on more than one background
Semi-transparent anti-aliased edges blend with the background. Look for halos around small letters and rounded shapes. The transparent emblem in our downloadable examples has partially transparent edges so the effect is visible. Check the downloaded file in the destination app because a checkerboard preview alone cannot confirm that app supports transparency.
An optimized source may not become smaller
If the input is already compact, another encode can be the same size or larger. MallowDesk keeps the original only when format and dimensions match and its metadata scan confirms no metadata to remove. If metadata is present or uncertain, the app uses the new encoding even when it is larger. The result label explains that tradeoff.
Try it with a reproducible example
Open the source files, measured outputs, and settings. Use the same source for each comparison and inspect the actual result.