Load Image (Inspire)

You know how you download someone's cool img2img workflow, drag it in, and the input image slot is empty because the picture lived on their machine and never came with the JSON? Load Image (Inspire) is the fix for that from the other direction. It works like the normal Load Image node, except the image itself gets stored inside the workflow. Share the workflow and the reference image goes along for the ride - no "where's the input?" on the receiving end.

The pack's own framing sums it up: the loaded image is stored in the workflow, making it easier to reproduce a generation on another computer. That's the entire pitch, and for anyone who shares or archives workflows it's a real convenience.

How it works

Instead of pointing at a filename in your input folder and hoping the recipient has the same file, this node embeds the image data (base64) directly into the node's saved state. When ComfyUI serializes the workflow to JSON, the picture is part of it. Load that JSON anywhere and the image reconstitutes itself. It also hands you the alpha channel as a mask, same as the standard loader, so it drops straight into inpainting and img2img graphs.

The obvious trade-off: your workflow JSON gets bigger - sometimes a lot bigger - because it's now carrying a whole image. For a small reference that's nothing; for a 4K input it's noticeable. That's the deal you're making in exchange for self-containment.

The inputs and outputs

• image - the image selector. It carries the picker plus a #DATA entry that represents the embedded-data path; in practice you load an image the usual way and the node stores it.
• image_data (STRING) - where the embedded image data itself lives. You don't hand-edit this; it's populated when the image is stored.
• Outputs are IMAGE and MASK (the alpha), exactly like the stock loader, so it's a drop-in replacement anywhere Load Image would go.

How to install it

ComfyUI Manager: Install Custom Nodes → search ComfyUI Inspire Pack → install → restart. Or clone it:

cd ComfyUI/custom_nodes
git clone https://github.com/ltdrdata/ComfyUI-Inspire-Pack

Then restart. The pack is Dr.Lt.Data's - the author of ComfyUI Manager and Impact Pack - so it's well-maintained and installs without drama. No model downloads for this node.

Common issues

The main thing to internalize is the file-size cost. If you're batch-sharing dozens of workflows or working with large inputs, embedding every image makes the JSONs heavy; use the regular Load Image for throwaway local work and save this for workflows you actually intend to hand off or archive. Second, because the image lives in the workflow rather than your input folder, editing it means re-loading through the node - there's no file on disk to swap. And if a shared workflow using this node opens with an empty slot, it was likely saved before the image was embedded; the original author has to re-save once the picture is in.