Choose output by image content
Conversion is not simply a request for the smallest possible file. First identify what visual information must remain. A transparent PNG logo, icon, or illustration needs an alpha channel; JPEG has none and replaces transparent pixels with a visible background. Keep PNG for simple graphics, or test WebP when you require both transparency and lower weight. A photograph without transparency can often be delivered as JPEG or WebP. Keep the source file: a web thumbnail, email attachment, and print image have different quality, size, and compatibility requirements. The right format preserves the necessary pixels at an acceptable size for its actual destination.
Test a transparent graphic and a photograph
For a transparent PNG, note width, height, and bytes. Export PNG without resizing, then WebP. Put both outputs over white and dark backgrounds to find opaque rectangles, halos, or damaged shadows. Do not use JPEG when transparency matters; if JPEG is required, choose its background deliberately. For a photograph, choose a copy with gradients, dark areas, foliage, and small text. Export JPEG or WebP at a middle quality setting and compare at 100 percent, not only as a thumbnail. Inspect skin, lettering, straight lines, and sky transitions, confirm dimensions in another viewer, and record source and output bytes. A smaller output only helps if it is suitable at the real display size.
Verify five properties before delivery
Check width and height and keep aspect ratio unless a crop is intentional. Test alpha against contrasting backgrounds. Compare color and contrast with the original on a reasonably calibrated display. Zoom into fine edges, then view the result at the size visitors will use. Finally compare bytes, but reject a saving that makes text fuzzy or introduces artifacts. JPEG and many WebP settings are lossy; repeated exports compound damage, so produce final variants from the original rather than reconverting an old output. Canvas conversion can omit or alter EXIF orientation, GPS, dates, color profiles, and copyright metadata. Inspect the result; do not assume metadata was removed or preserved.
Browser limits and supported files
Local conversion can hold the source, decoded pixels, canvas, and encoded output in memory simultaneously. A high-resolution photograph may need much more memory than its disk size suggests, especially on mobile devices. Test a copy first, reduce dimensions before large batches, and stop if the browser becomes unresponsive. This tool supports JPG, PNG, and WebP. HEIC, TIFF, RAW, layered files, and animated SVG can require specialist software or a separate workflow. Use the linked image optimization guide for delivery decisions and the EXIF privacy guide when metadata must be inspected after conversion.
Keep a reproducible record
Store the original alongside each delivered variant and record the output format, dimensions, quality setting, and final byte size. That record makes it possible to repeat a conversion when a layout changes or a colleague reports a visible problem. A browser preview is useful, but it is not a complete check: download and open the final file. Where an image has editorial, legal, or commercial value, keep an untouched source and check its color profile in the software used by the recipient. Conversion simplifies delivery; it does not replace responsible archiving, visual review, or rights management.
A short delivery checklist
Before sharing a converted image, open the downloaded file rather than relying only on the preview. Confirm its width and height in pixels, its aspect ratio, and the format shown by a second viewer. Place transparent artwork over both a light and a dark background. For a photograph, compare the original and export at the intended display size and at full size; compression defects often appear only in gradients, foliage, lettering, or high-contrast edges. Note the output byte size and reject a smaller file if it harms the purpose of the image. Keep the original and the settings used. This simple record makes a later correction possible without repeatedly recompressing an already lossy copy.