Duplicate WebP

No. The HTML5 Canvas API strips all metadata – EXIF, color profiles (ICC), and any custom chunks – when you draw an image onto a canvas. This is a privacy/security default of the spec, not specific to this tool. Compression settings, quality level, and metadata are likewise not carried over. If you need metadata-preserving WebP tiling, use a desktop tool.

Image loading is asynchronous: reading the width right after setting img.src returns 0, which would size the canvas to 0 × 0 and produce blank tiles. The tool waits for the decode to finish, then cuts each tile from the real dimensions.

Safari versions before 14 (and a few older browsers) silently fall back to PNG when you request toDataURL('image/webp'). The tool detects this by inspecting the returned data URL’s prefix and reports the actual format in the stats line. The download filename uses the real extension (.webp or .png) so you never end up with a mislabeled file.

Yes, both PNG sources and WebP sources with alpha channels keep their transparency through the duplication. The canvas API preserves alpha when drawing image-to-image, and WebP encodes alpha when present.

The Clipboard API only reliably accepts image/png across browsers. So when you copy, the tool encodes a PNG from the canvas and writes that to the clipboard. The Download button still produces a .webp file (if the browser supports WebP encoding) – only the clipboard path is PNG.

Output is exactly source-width × cols by source-height × rows. A 2000 × 2000 source at 10 × 10 grid = 20,000 × 20,000 = 400 megapixels – likely too big for many browsers (canvas has implementation-specific max dimensions, often around 16,384 px per side and ~268 MP total). The tool warns via toast when projected output exceeds 50 megapixels.

No. The page loads three static files (HTML, CSS, JS) and then runs entirely in your browser. You can disconnect from the internet after the page loads. No analytics, no tracking, no cookies.

Yes – free, unlimited, no signup, no watermark, no usage cap. Use the output in any context. Attribution to is appreciated but not required.