Base64 to Image

Processing pipeline, step by step: as you type or paste, input is debounced by 300 milliseconds (a preview re-renders 300ms after your last keystroke, rather than on every individual character) so pasting a very long string doesn't trigger dozens of redundant decode attempts in quick succession. When the debounce timer fires, the current textarea content is stripped of all whitespace — spaces, tabs, and newlines alike — using a single regular-expression replace (`input.replace(/\s+/g, '')`), which fixes the exact copy-paste artefacts described above: wrapped lines, stray leading/trailing spaces, and any accidental extra whitespace picked up in transit. After trimming, the tool checks whether the resulting string already begins with `data:` — if it does, it's used exactly as-is, on the assumption that whatever prefix is present is the correct, intended one. If it doesn't start with `data:`, the tool builds a data URI itself by prepending `data:image/<assumedFormat>;base64,` in front of the trimmed text, where `<assumedFormat>` comes from the format select — defaulting to `png` — that you can change to `jpeg` or `webp` at any time, which immediately re-triggers the same decode logic with the new prefix. The actual decoding is delegated entirely to the browser: the constructed (or as-is) data URI is set directly as an `<img>` element's `src` attribute, and the browser's own native image decoder does the work — this tool contains no custom Base64-to-pixel decoding logic of its own, which is both simpler and more reliable than reimplementing image format parsing in JavaScript. Two events on that `<img>` element drive the rest of the interface: `load`, which fires when the browser successfully decodes and renders the image (this both clears any previous error and enables the download button), and `error`, which fires when the browser cannot make sense of the data — either because the Base64 itself is corrupt/incomplete, or because it decoded as bytes that simply aren't a valid image in the format claimed by the prefix. On an `error` event, the tool shows the message "This doesn't look like valid image data" rather than leaving a broken-image icon on the page, and disables the download button until a subsequent edit produces a successfully decoded preview. Downloading uses the simplest possible approach available: since the currently-previewed data URI is already a complete, self-contained representation of the image, the download button simply creates a temporary `<a>` element with its `href` set directly to that data URI and a `download` attribute set to a filename built from the detected image type (extracted from the `data:image/<type>;base64,` prefix, with `jpeg` mapped to the more conventional `.jpg` file extension), then triggers a click on it — no intermediate Blob construction or conversion step is needed, since the browser already understands how to save a `data:` URI directly to disk via the `download` attribute.