IMAGE Web Viewer Channel Loader @ vrch.ai

The Web Viewer nodes are sinks - they take an image and show it in a browser. The IMAGE Web Viewer Channel Loader is the tap back into the graph: it pulls the image that's currently being displayed on a given viewer channel and hands it to your workflow as a normal IMAGE. You can save it, run img2img on it, upscale it, process it, loop it back around. Whatever the viewer is showing, this node can grab it.

What it is

The mechanism is the pack's channel system. Every viewer node in ComfyUI Web Viewer has a channel setting from 1 to 8, and this loader has the same one. Point the loader at channel 1, point a viewer at channel 1, and the loader outputs whatever that viewer is currently displaying. It's how a realtime setup stays consistent across the pack: the viewers share a little channel registry, and this node reads from it. The single channel input is genuinely the only knob - pick a number, match it to a viewer, done. The output is a single IMAGE.

Where this earns its keep: archiving. In an instant-queue interactive workflow you're generating constantly, and the default Save Image node names files in a way that's hard to track. Chain this loader into the pack's IMAGE Saver with your own filename scheme, and you capture exactly what the audience saw, when they saw it. It's also the natural input for a feedback loop - grab the current frame, drop its resolution, feed it back into the sampler for the next pass. Real-time looping pipelines are the pack's whole genre, and this is the node that makes them possible.

It's also a gentle on-ramp to the pack's architecture. Once you understand channels, you understand why there are so many "channel" fields scattered across these nodes: the viewer, the loader, the preview nodes all coordinate through that number. Channels let you run several independent display pipelines in one workflow without them colliding - viewer A on channel 1, viewer B on channel 2, and this node can read either without caring which workflow produced it.

How to install

Install is the pack-standard one. Search "ComfyUI Web Viewer" in ComfyUI Manager, or:

cd ComfyUI/custom_nodes
git clone https://github.com/VrchStudio/comfyui-web-viewer
cd comfyui-web-viewer
pip install -r requirements.txt

then restart. Windows portable users run the pip line through python_embeded\python.exe. No models or heavy dependencies - this is a small bridge node.

Common issues

The main gotcha is channel mismatch: if nothing comes out, check that the loader's channel matches a viewer's channel, and that the viewer has actually displayed something. A channel that's never been used has nothing to load. It's also worth noting the loader is one of the nodes that makes sense to pair with a fallback - a realtime display might not have produced a frame yet when the graph first runs, and the IMAGE Fallback node from the same pack will happily hold the fort until it does.