Load Image From Path
Load an image from any path — or a URL — not just the input folder
- IMAGE
- MASK
ComfyUI's built-in LoadImage only sees files inside your input folder. JDC_ImageLoader throws that rule out: you type any absolute path - or even an http(s) URL - and it loads whatever image you point at. It's the node you want when your source images live somewhere else entirely, or when a workflow hands you a path as text and you want it loaded without copying files around.
It's the plain loader from comfy-plasma by Jordach, and it does the job with some thoughtful touches under the hood.
How it works
The single input is image, a string. The loader does a few quiet things before Pillow opens the file:
- Strips surrounding quotes. If you copy a path from Windows Explorer ("Copy as path"), it comes with
"quotes attached. This node removes them automatically, so pasted paths just work. - Handles URLs. If the string starts with
http, it fetches the image withrequestsand loads it from memory. No saving to disk first. - Auto-fixes rotation. It applies
exif_transpose, so phone photos with EXIF rotation flags come in upright instead of sideways.
It returns two outputs: IMAGE (the pixels) and MASK. The mask comes from the image's alpha channel if it has one (inverted, so transparent = 1, the ComfyUI convention). No alpha? You get a small 64×64 zero mask, which is the standard "no mask" sentinel.
The nice engineering bit: it implements IS_CHANGED by hashing the file's contents. That means if you re-run your workflow after editing the image on disk, ComfyUI re-loads the fresh version instead of serving a stale cache. And VALIDATE_INPUTS gives you a clear "No file found: /path/you/typed" error the moment you queue a run with a typo, rather than a cryptic traceback.
When you'd reach for it
- Your reference images live in another drive or a shared folder and you don't want to duplicate them into
input/. - Automation: a workflow that takes a path as a text input (or from another node) and loads whatever it's told.
- Quick tests against files you're actively editing - the content-hash caching means it always picks up the latest save.
Install
No requirements.txt, no models. ComfyUI Manager → Install Custom Nodes → search comfy-plasma → install, restart. Or:
cd ComfyUI/custom_nodes
git clone https://github.com/Jordach/comfy-plasma
Restart ComfyUI. It's under image.
Common issues
- Path typos - you'll get the friendly "No file found" validation error. Double-check slashes on Windows; backslashes in strings can be a nuisance, and it's exactly why the quote-stripping exists.
- URLs need network - a remote image that's moved or blocked just fails to open. Nothing cached to disk, so it re-fetches every run (which also means slow runs if the URL is slow).
- The mask output is only meaningful for images that actually have an alpha channel (PNG/TIFF with transparency). A JPEG gives you the zero sentinel mask - don't go hunting for a "real" mask where there isn't one.
Inputs (1)
| Name | Type | Default | Description |
|---|---|---|---|
| image | STRING | — |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| IMAGE | IMAGE | — |
| MASK | MASK | — |