H3 Continuum Load Image

Look at the H3 Continuum Sampler V3.8 and you'll see five still-image sockets staring back at you: first_frame, last_frame, and reference_image_1 through reference_image_3. This node is the file picker you hang off all of them. It's ComfyUI's standard Load Image wearing an H3 label and carrying one genuinely useful trick: an Enable Image switch that bypasses the node without you ripping out the cable.

That matters more than it sounds. Continuum's own docs lean hard on one reusable template that exposes every optional slot - First Image, Last Image, and three Reference Image slots - so you can flip features on and off per run without editing the graph. With a plain Core Load Image you'd be right-clicking and choosing "Bypass" or deleting connections constantly. With this node, Enable Image OFF behaves exactly like an unconnected optional input, and ON runs the normal loader. Same node, five copies, zero graph surgery.

What it is under the hood

H3EasyLoadImage subclasses ComfyUI Core's LoadImage - the Enable Image toggle is frontend-only JavaScript that flips the node into ComfyUI's native bypass mode. You get the exact same IMAGE / MASK contract as the stock node, so anything the H3 sampler accepts from Core LoadImage, it accepts from here. One image file dropdown in, IMAGE and MASK out.

Why you'd bother

The Continuum sampler does its own image sizing now. In V3.8 the Size Source = First Image mode takes its aspect ratio from the connected First Image and applies one of the resolution presets, and Reference Images are resized internally before VAE encoding. So you don't need an image-resize node upstream - you just feed this loader's IMAGE straight into whichever socket fits the job, and the sampler handles geometry. Drag the same node in for First, Last, and each Reference; use the toggle to switch off the slots you don't need this run.

Installing

These nodes ship together in one pack, so the install is a single clone:

cd ComfyUI/custom_nodes
git clone https://github.com/ukr8b3g-cmyk/ComfyUI-H3-Continuum.git

Restart ComfyUI after cloning (or search "H3 Continuum" in ComfyUI Manager and let it handle it). There are no extra pip dependencies - the pack rides on PyTorch and safetensors ComfyUI already bundles, and needs ComfyUI 0.32.0 or newer.

Gotchas worth knowing

If the Enable Image widget doesn't appear, the Python-defined node interface still loads and runs - the toggle is an extension, not the loader. One bigger caveat that has nothing to do with this node: it loads whatever image you point it at, but the model it feeds is another matter. MiniMax H3 weights come from HuggingFace under the MiniMax H3 Community License, which excludes the US, EU, UK and South Korea from its Applicable Territory - worth checking before you build a whole workflow around it. And remember the H3 Continuum pack bundles no models, LoRAs, images, or audio; everything you actually generate with has to come from you.