RedNode Rig Result

A node with zero outputs that renders nothing itself is the kind of thing that makes people assume they wired it wrong. You didn't. RedNode Rig Result is a full stop, and the fact that nothing leaves it is precisely the point.

Why it exists

RedNode Studio's Workspace normally samples internally - one panel drives the model, the prompt, the latent and the sampler, and hands out a finished picture. When a rig is set to kind Your own nodes, the pack has to run your sampler somewhere in the middle, with the Workspace's conditioning and numbers as inputs, and then pick the picture back up and keep going: Detailer passes, Post FX, Save, the review strip.

That sandwich needs two hands. RedNode Rig Inputs is the Workspace handing you the prompt, latent, seed and sampler settings. Rig Result is the other hand: your chain puts the finished thing down here, and the Workspace takes it from there.

How it works

Rig Result is the anchor the pack searches for. If the queued prompt contains no Rig Result naming the rig in use, your whole chain is skipped and the rig samples with the built-in sampler instead - a silent-ish fallback that is worth knowing about, because it produces a plausible image from settings you didn't choose.

When it does find one, the pack walks backwards from this node along the links, stopping when it hits Rig Model and Rig Inputs, and runs every node in between using ComfyUI's own node-calling machinery. That has a pleasant side effect: nodes that don't depend on the rig's values - your loaders, a model patch, a constant - are detected as independent and reused across calls, so your checkpoint loads once per queue even if the rig is sampled five times.

And it will be sampled more than once. The main render, every Latent refine pass, the Img2Img pass, the Paint tab's render, and each Detailer pass on that rig all route through your nodes, one call at a time, each with its own steps, CFG, denoise and step window. So your chain is not a graph that runs once; it's a function the Workspace calls with different arguments.

The two inputs

Only one is required: rig, the string that has to match the other two rig nodes and the Models tab entry.

Then pick your ending, and pick the latent one if you can:

• latent - your sampler's output. This is the preferred route: the Workspace decodes it itself, so the pixels never round-trip through a VAE encode, and the pipeline carries on into the Detailer, Post FX and Save exactly as if the built-in sampler had produced it.
• image - a finished picture, for chains that decode inside themselves. The Workspace encodes it back with this rig's VAE so the next stage has a latent to work on. That's a lossy round trip you're paying for twice, and it drops the incoming noise_mask. Use it when your chain genuinely has to decode, not as the default.

Wire one or the other. Wiring both is a coin toss; wiring neither is an error with a clear message.

Install

Rig Result ships in the pack and needs nothing installed.

cd ComfyUI/custom_nodes
git clone https://github.com/RedNodeAI/ComfyUI-RedNodeStudio.git ComfyUI-RedNodeStudio

Or search RedNode Studio in ComfyUI Manager and restart. The three rig nodes appear under RedNode/Rigs.

Where people get burned

"The Rig Result for 'My rig' has no latent or picture wired into it." That's the pack telling you your sampler's output never made it here - usually because you rewired the sampler and left the old link dangling.

"The rig 'X' returned a picture, and there is no VAE to turn it back into a latent." The image route needs a VAE on the rig. Wire one into RedNode Rig Model rather than a separate VAE node - the rig reads its VAE from its own head node.

Dropped out of the chain. Bypass or mute a node between your sampler and Rig Result and the walk-back fails with a message naming the node as "not in this queue". Bypass Rig Result itself and the pack stops looking entirely: your sampler never runs and the built-in sampler quietly renders instead.

Two Rig Result nodes with the same rig name. The pack uses the first one it finds and warns you. In a graph you've copied rigs around in, that warning is the only thing keeping you from debugging the wrong branch.

One last disambiguation, because the names are close: the pack also ships RedNode Rig Out and Rig In, which hand a rig's prompt, seed and sampler numbers to an outside engine over actual wires. These three nodes are the no-wires version - the rig name does the connecting, and the Workspace keeps the canvas clean.