ICO to PNG Converter

This tool pulls the images out of a Windows .ico file and saves each one as its own PNG. An .ico is a container that can hold several icon images — usually different sizes like 16 × 16, 32 × 32, 48 × 48, 256 × 256 — and sometimes different colour depths. The tool looks inside the .ico, lists every image it finds, and extracts each one. For PNG entries stored inside the .ico (a feature added in Windows Vista) the original PNG bytes are copied out exactly, no re-encoding. For older-style bitmap entries, the browser decodes each bitmap and saves it as a PNG — a small, honest re-encoding by the browser.

A browser would usually decode an .ico as a whole and show only one of its frames. To give you every size, the tool parses the .ico's directory and, for each entry, either slices out the stored PNG or builds a single-entry .ico around that bitmap and lets the browser decode that single image. The result is one PNG per directory entry, named after the icon and its size (e.g. `icon-16x16.png`, `icon-256x256.png`). Windows stores a 256 px image as 0 in the one-byte width/height fields; the tool uses 256 for naming and display.

An .ico can hold any combination of PNG and bitmap entries — some real files mix them. The page is explicit about which is which: PNG entries are copied as stored (lossless), and bitmap entries were decoded by your browser and re-encoded. Transparency is preserved where possible. Bitmaps below 32 bits per pixel use a 1-bit AND mask for transparency; 32-bit bitmaps can carry an alpha channel.

Everything happens in your browser. Nothing is uploaded. The file is read locally, the directory is parsed, and output PNGs are created in-memory for you to download individually or as a ZIP.

1. Read the 6-byte ICONDIR: `reserved` (0), `type` (1 for icons), `image count` (uint16 LE). A count of 0 is refused. The count is also capped to stop an absurdly long list of images from exhausting memory in a browser tab. 2. Read one 16-byte ICONDIRENTRY for each image: `width` (byte; 0 means 256), `height` (byte; 0 means 256), `colour count` (byte), `reserved` (byte), `colour planes` (uint16 LE), `bits per pixel` (uint16 LE), `byte size of the image data` (uint32 LE), `offset of the image data` from the start of the file (uint32 LE). The tool uses the directory's width and height for naming and for rendering. 3. Check that the directory fits in the file, and that every entry's `offset + size` fits in the file, naming the first offending entry. Also reject entries that are neither a PNG nor a bitmap that begins with a known BITMAPINFOHEADER-family size. 4. Classify each entry: if the data at `offset` starts with the PNG signature (`89 50 4E 47 0D 0A 1A 0A`) → `kind: png` and the PNG bytes are copied out byte for byte. Otherwise, if the first four bytes form a recognised DIB header size (12, 40, 52, 56, 64, 108 or 124) → `kind: bmp` and the browser will need to decode it. Anything else is not image data and is refused. 5. For each entry, produce a PNG named `<name>-<w>x<h>.png`. For PNG entries this is lossless extraction. For bitmap entries the widget wraps only that entry in a minimal single-entry .ico (rewriting the data offset to the new position) and passes it to the browser's image decoder; the decoded bitmap is drawn onto a canvas at the entry's size and encoded as a PNG.