Nodes/SP-Nodes/GodnessMerger_Out
ComfyUI Node

GodnessMerger_Out

Merging the one layer that actually produces the image

By bananasss00·Created 2 years ago·Updated 6 days ago· 20
GodnessMerger_Out
    • GM_OUT
    seed0
    min0.0
    max1.0
    type
    custom_value1.00

    GodnessMerger_Out is another data-only piece of SP-Nodes' modular "Godness Merger" merge system - it takes no model and produces no model, just a ratio value tagged GM_OUT, meant for a combiner node elsewhere in the pack that this batch of briefs doesn't cover. If you're new to this family, start with whichever article walks through the overall idea; this one's short because the node itself is small, in every sense.

    Why "out" gets its own dial

    The Godness Merger nodes are named after the real sub-modules inside the classic Stable Diffusion U-Net class: time_embed, label_emb, middle_block, output_blocks, and out. out is the very last stage - a GroupNorm, a SiLU activation, and a final convolution that maps the network's internal hidden features back down to the latent's channel count. It's tiny relative to the rest of the model, but it's the literal last thing the U-Net does before handing a noise prediction back to the sampler. Giving it an independent merge ratio makes sense for the same reason time_embed and label_emb get their own nodes instead of folding into one blanket ratio: it's a small, structurally distinct piece that behaves differently from the big encoder/decoder stacks around it, so treating it separately gives you finer control than one global merge weight would.

    What it does

    Same mechanic as its siblings: seed, min, max, type (custom or random), custom_value. custom mode returns custom_value exactly. random mode draws a number between min and max, deterministically from seed - rerun with the same seed and you get the same ratio. That's it; the output is a number with a label on it, GM_OUT.

    The fields worth setting

    • type - custom if you want an exact ratio, random for a reproducible roll.
    • min / max - the 0–1 range random mode samples from.
    • custom_value - only used in custom mode.
    • seed - makes "random" repeatable, which matters if a random draw happens to land on something good.

    Installing SP-Nodes

    Manager: search "SP-Nodes," install, restart ComfyUI. Manual clone into your node folder:

    cd ComfyUI/custom_nodes
    git clone https://github.com/bananasss00/ComfyUI-SP-Nodes
    

    Restart after. Nothing about this node needs a model download - it's arithmetic on floats.

    Troubleshooting

    Standard missing-node red box if you load a workflow using this before installing the pack - install via Manager, restart. Beyond that, the one thing worth remembering: because out is such a small layer, don't expect merging it in isolation to shift your results as dramatically as merging middle_block or output_blocks would. If you're chasing a specific visible effect and tweaking this node's ratio isn't moving the needle, that's not necessarily broken - it might just be that the effect you're after lives in a bigger block. Confirm whatever combiner node you're using downstream is actually wired to consume GM_OUT at all before assuming the ratio is the problem; a disconnected data node fails silently in ComfyUI rather than throwing an error, which is the most common way people lose an hour chasing a "why isn't this doing anything" that turns out to be a missing wire.

    CategorySP-Nodes/model_merging

    Inputs (5)

    NameTypeDefaultDescription
    seedINT00–18446744073709550000
    minFLOAT0.00–1
    maxFLOAT1.00–1
    typeCOMBO2 options: custom, random
    custom_valueFLOAT1.000–1

    Outputs (1)

    NameTypeDescription
    GM_OUTGM_OUT