Siberia 图片加载器 / Image Loader

Let's be honest up front: the Siberia Image Loader is the core LoadImage with one thing bolted on - a second output that tells you whether the load actually worked. If that sounds marginal, you're right, and that's the point of the node. It exists to make the pack's own Ollama Vision workflow easier to debug, and its one real trick is that it fails loudly instead of silently.

What it actually does

You get a single required input, image - a dropdown of every image file in ComfyUI's input folder, with the little upload button on the widget so you can drag a file straight in. Pick one and it comes out two ways:

• 图片 / Image (type IMAGE) - the standard image tensor, converted to RGB, normalized to 0–1 floats, batch dimension added. Identical pipeline to the core loader.
• 信息 / Info (type STRING) - a status line. On success it's the full path, dimensions and mode of the loaded file; on failure it's an error string instead of a crash.

That second socket is the whole product. Wire it into the pack's Universal Display node and you can see "Successfully loaded image: …" or "Error loading image: File not found" on the canvas, which beats staring at a black preview and guessing.

Where you'd actually reach for it

Inside this pack, feeding the Siberia Ollama Vision node - the loader's 图片 / Image output plugs straight into the vision node's images input. If you're already running the pack, you might as well use it; the info string is free diagnostics. Outside the pack, the core LoadImage is just as good, so don't install this package for the loader alone.

The trap to know about

When a load fails, the node returns a 1×64×64×3 tensor of zeros - a real, image-shaped black frame - alongside the error string. Downstream nodes won't raise; they'll happily feed a black square to your vision model and you'll get a confident description of nothing. That's exactly why the 信息 / Info output matters: it's the only place the failure is visible. Check it whenever a vision result looks off.

Install

It's one of six nodes in the ComfyUI-SiberiaNodes pack, so you get it with the whole package:

cd ComfyUI/custom_nodes
git clone https://github.com/siberiah2o/ComfyUI-SiberiaNodes.git
cd ComfyUI-SiberiaNodes
pip install -r requirements.txt

Restart ComfyUI and it's under Siberia Nodes/Image. ComfyUI Manager users can just search "ComfyUI-SiberiaNodes" and hit install. The dependencies are light - torch, numpy, Pillow are already in your ComfyUI environment; the pack only adds the Ollama client, aiohttp and PyYAML.

Troubleshooting

• File not in the dropdown? Use the upload button on the widget, or drop it into ComfyUI/input and refresh. The dropdown only lists files already in that folder.
• Output is black and you don't know why? Read the 信息 / Info string. If it starts with "Error", that's your cause, not your tensor.
• Trust it? It's a small, essentially unknown pack (no community footprint worth mentioning), so glance at utils/image_loader.py once before you run it. What I found when I read it: it touches only your input folder and your configured Ollama server - no API key, no surprise phone-homes.