Nodes/comfyUI-tool-2lab/show web image 显示网图
ComfyUI Node

show web image 显示网图

Load an image from a URL as an actual tensor

By AI2lab·Created 2 years ago·Updated 2 years ago· 23
show web image 显示网图
    • image
    • mask
    • image_url
    • filePath
    image_url
    RGBAfalse

    This is the sleeper hit of the pack. show_web_image (2lab) takes a URL, downloads the image, and turns it into a real IMAGE tensor you can wire into a sampler, a VAE encode, an IP-Adapter, whatever. Core ComfyUI's Load Image only reads from disk - there's no built-in "load from the web." This node is that missing piece, and it's genuinely useful in more workflows than the pack's otherwise quiet reputation suggests.

    Why bother? Because a growing share of real workflows start with a reference image that lives behind a URL: a reference from a gallery, a frame from a video service, an image your own API returned. Before this node, you'd download it with a browser, find it on disk, and then load it - a manual detour every single time. With this node, the URL flows straight into the graph.

    Inputs

    • image_url - the URL of the image to fetch. Multiline not needed; it's a single link.
    • RGBA - a toggle (default false). Flip it true when you need the alpha channel preserved. When you do, the node also hands you the image's alpha as a MASK, which is how you composite transparent web images (PNGs with alpha, logos, stickers) onto a background without losing the transparency.

    Outputs

    • image - the downloaded image as a proper IMAGE tensor. This is the one you'll wire onward.
    • mask - alpha channel as a MASK, populated when RGBA is on.
    • image_url - the URL echoed back as text, handy for bookkeeping.
    • filePath - where the fetched file landed locally. Useful if a downstream node wants the path rather than the tensor.

    How it works

    Fetch the bytes, decode them into a tensor, and - because this pack is built around the 2lab server - it caches the file to disk so a published workflow doesn't re-download the same image on every call. The filePath output is that cache location. If you only care about the tensor, you can ignore everything except image.

    Installing it

    Same install as the whole pack. ComfyUI Manager → search "comfyUI-tool-2lab" → install, or:

    cd ComfyUI/custom_nodes
    git clone https://github.com/AI2lab/comfyUI-tool-2lab
    

    Restart ComfyUI. No model downloads.

    Gotchas

    • Network dependence is the whole risk. A dead link, a site blocking hotlinks, or a redirect the fetch doesn't follow means the node fails and your queue stops. Validate URLs before you wire them in.
    • RGBA on, mask follows. If you need the alpha, remember the mask only appears when RGBA is true. Leave it off and you get an opaque image - fine for reference work, wrong for compositing.
    • Repo visibility. As of this writing the pack's GitHub page returns 404 - likely made private or renamed. A fresh clone can fail even though the node still shows up in shared workflows and installed setups. Don't read that as your install error.
    Category🦊2lab/util/preview

    Inputs (2)

    NameTypeDefaultDescription
    image_urlSTRING
    RGBACOMBOfalse2 options: false, true

    Outputs (4)

    NameTypeDescription
    imageIMAGE
    maskMASK
    image_urlTEXT
    filePathfilePath