Simple Web Browser

Ever catch yourself alt-tabbing out of ComfyUI to hunt for a reference image, save it to disk, then drag it back in? That dance is the whole reason this node exists. Simple Web Browser drops an actual browser - an iframe, technically - right inside a node on your canvas. You browse for inspiration, and when you find the image you want, you click a button and it shows up on the node's output as a ready-to-wire IMAGE tensor. No save, no drag, no leaving the graph.

It's the same author's family of "asset gallery" packs (Civitai Gallery, Local Image Gallery), and the framing is the same: stop treating ComfyUI and the web as two places. The reddit reception was small but warm - it's a niche convenience, not a heavyweight tool, and it knows it.

How it works

There are two halves. A JavaScript extension builds the browser widget: an address bar, a 🏠 home button (defaults to Photopea, right-click to set your own), a bookmark menu saved to browser_bookmarks.json in the plugin folder, and an <iframe> doing the actual browsing. Underneath, a tiny Python node (WebViewerNode) waits to turn what you load into tensors.

Here's the part that trips people up: the node has no visible inputs at all. Check the info schema and required is empty - the image URL is passed through a hidden widget, and the node just runs. So don't go hunting for an input socket to connect; you set it from inside the browser UI.

You can load an image three ways:

• From URL - paste a direct image link (.jpg, .png, etc.) and hit Loading pictures from url. The fetch happens server-side with Python requests, so CORS isn't your enemy here.
• From clipboard - paste an image straight into the input box; it's uploaded to a /browser/upload_temp_image endpoint and dropped in ComfyUI's temp directory.
• From Photopea - the genuinely clever one. Browse to Photopea, edit your reference image on the canvas, hit From Photopea, and it exports the canvas via postMessage and pulls it in. A full online image editor feeding your graph directly.

The outputs that matter

• image (IMAGE) - the loaded image, with any alpha composited onto black so downstream nodes see a clean RGB tensor.
• mask (MASK) - the alpha channel. If you loaded a transparent PNG, that's the transparency, ready to feed an inpaint mask or a compositing node. That pairing is the most useful part of this node: a transparent reference comes in with its cutout for free.

Because the node keys its IS_CHANGED off the image URL, loading a new image re-executes it automatically. Note that it has to actually run - the output isn't live until the workflow executes.

Installing it

ComfyUI Manager → search ComfyUI_Simple_Web_Browser, or clone it:

cd ComfyUI/custom_nodes
git clone https://github.com/Firetheft/ComfyUI_Simple_Web_Browser

Then restart ComfyUI. There's no requirements.txt, no pip install, no model downloads - it uses only torch, numpy, requests, and PIL, all of which ComfyUI already ships. Restart matters because it's a front-end node (WEB_DIRECTORY), so a fresh browser reload is what actually loads the widget.

Where you'll get burned

• A lot of sites just won't render. The iframe is at the mercy of X-Frame-Options and CSP, so GitHub, Google, and plenty of big-name sites either refuse to frame or break. The README is upfront about this. Treat it as "see which of your favorite sites work," not as a Chrome replacement. Image hosts and lightweight sites usually behave; Photopea clearly does.
• The URL has to be a direct image link, not a page you're viewing. And because the fetch only sends a Mozilla User-Agent with no referer, hotlink-protected hosts will 403 you. Find the raw image URL, and if a host blocks it, that's the host, not you.
• Clipboard paste lands in the temp folder, keyed by filename - it works, but don't build a pipeline that expects those files to live forever.

It's not a daily driver for everyone, but if you spend real time hunting reference images, the Photopea round-trip alone is worth the install.