JPG, PNG, and WebP: which format to export for thumbnails, posts, and screenshots

Thumbnail, banner, Story, and blog hero are not the same encoding problem. The compress & convert and resize & crop tools on this site re-encode with the browser Canvas API. Canvas does not keep a layered PSD. It paints pixels and writes a new file. Choose the container on purpose.

JPG (JPEG)

Best default for photographs and YouTube thumbnails that are already busy pictures. You pick a quality slider; artifacts show up as muddy edges and banding in skies. YouTube’s 1280×720 thumbnail is a photo-shaped job — JPEG at a high quality setting is usually enough. Do not JPEG a screenshot of UI chrome or text-only graphics if you can avoid it; the letters will fuzz.

PNG

Lossless for the pixels it stores, and it can hold alpha (transparency). Use it for logos, simple graphics, and screenshots you still need to read. PNG can be huge. Our compressor offers scale presets because “PNG quality” is not a JPEG-style slider — shrinking dimensions is the reliable way to cut weight. If you flatten a transparent logo onto a thumbnail, you no longer need PNG.

WebP

Often smaller than JPEG at a similar look, with optional transparency. Many modern sites accept it. Some older uploaders, email clients, and a few social forms still do not. If a platform rejects the file, convert to JPEG or PNG and stop fighting the CMS. WebP is not a moral upgrade; it is a compatibility bet.

Canvas WebP support depends on the browser you are in. If export fails, try Chrome or Firefox, or pick JPEG. We cannot polyfill a missing encoder.

Resize before you argue about format

A 4000-pixel phone photo uploaded as a 1280×720 thumbnail wastes bytes twice: once in the file, again when YouTube transcodes it. Crop to the aspect you need (16:9 thumb, 16:9 banner with safe area, 1:1, 9:16), then encode. The resize tool’s cover mode is “fill the frame and cut the rest.” Contain letterboxes. They are different jobs; preview and export share the same geometry so you are not surprised.

Batch caps (roughly 20–24 images) exist because decoding many large photos in one tab is how browsers die. That is not an upsell.

Metadata is a separate pass

Camera JPEGs often carry GPS and device strings. Metadata Cleaner re-encodes images through Canvas so EXIF is dropped. Compress/convert also re-encodes, so it usually drops EXIF as a side effect — but if privacy is the point, use the cleaner and check the download. PDFs and audio have different metadata paths (pdf-lib info fields; FFmpeg -map_metadata -1).