Load Image (Path) 🎥🅥🅗🅢

ComfyUI's built-in Load Image node makes you upload a file and pick it from a dropdown. Sometimes you don't want that - you want to load a single image straight from a filesystem path, maybe one produced by an earlier step or living outside the input folder. That's what this node is for. It's the single-image, path-based loader in Kosinkadink's VideoHelperSuite, and it adds two conveniences the native loader lacks: on-the-fly resizing and an optional VAE that turns the image into a latent as it loads.

It's a small, unglamorous utility, but it earns its keep in automated and multi-stage graphs where the image you want isn't sitting in the input dropdown - think a start frame for a video, a reference pulled from a working directory, or a file a script just wrote.

How it works

You give it a path, it decodes the image, and optionally rescales it before handing it off. If you connect a VAE, it encodes to a latent on the spot rather than making you add a separate VAE Encode node. Nothing fancy - it's the missing "load from path" primitive.

The inputs and outputs that matter

The image widget is the path. It's a VHS_PATH field with completion - type, Tab to accept, Ctrl+B to go up a directory. The resize knobs:

• custom_width and custom_height - target dimensions, in steps of 8. Set both and it crops to maintain aspect ratio; leave at 0 to skip resizing entirely.

The optional vae is the interesting one: wire in a VAE and the node outputs a latent instead of a plain image, saving you a node.

Outputs: IMAGE and mask (the alpha channel, if the file has one).

How to install it

ComfyUI Manager: search ComfyUI-VideoHelperSuite, install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite

then restart. No model downloads.

Common issues & troubleshooting

Path won't resolve. On hosted setups, VHS_STRICT_PATHS can restrict loading to ComfyUI's directories. If it's set, drop the file under ComfyUI's input folder and use the built-in Load Image or the upload nodes instead.

The mask is blank. Your image has no alpha channel - normal for jpg and most png exports. The mask only carries data if transparency exists in the file.

The resize distorted my image. With only one of custom_width/custom_height set it scales to that dimension; set both to crop-to-fit, or leave both at 0 and resize with a dedicated node downstream if you want more control.