flow_bridge_image

No description shipped with this one either, but its four parameters lay out the job clearly enough: it's a pass-through checkpoint for an image and mask pair, with independent kill switches on the input side and the output side. Think of it as a valve you drop into a graph so you can cut a section out of the execution path - for A/B comparisons, for debugging a subgraph, for temporarily disabling a heavy branch - without unplugging any wires.

How it works

image and mask come in as optional inputs and go back out unchanged as image and mask outputs - a straight bridge, by default. disable_input and disable_output are the two toggles: flip disable_input and whatever's upstream stops mattering to this node; flip disable_output and this node stops pushing its result downstream. Having both as separate switches (rather than one "bypass" flag) means you can, for instance, keep the input side live so an expensive upstream node still runs and caches its output, while cutting the output side so nothing downstream actually consumes it yet - useful when you're iterating on a later part of a graph and don't want to keep re-running an earlier part. It's flagged as an output node, so it's also valid to leave it as a dead end in your workflow.

Inputs and output

• disable_input (BOOLEAN, default false) - cuts what flows in.
• disable_output (BOOLEAN, default false) - cuts what flows out.
• image (IMAGE, optional).
• mask (MASK, optional).
• image / mask (outputs) - the bridged values, when both switches leave them enabled.

Installing it

Search ComfyUI-Apt_Preset in ComfyUI Manager and install, or manually: cd ComfyUI/custom_nodes && git clone https://github.com/cardenluo/ComfyUI-Apt_Preset.git, then restart ComfyUI. The README's dependency step is "double-click install.bat" - a Windows-first instruction, with no requirements.txt visible for other platforms, so on Linux/Mac open that file to see what it installs and replicate it manually in your ComfyUI venv. This node is pure graph plumbing, so it needs none of the pack's optional extras (GGUF, Advanced-ControlNet, nunchaku, downloaded models).

Common issues

Both inputs are optional, so it's entirely possible to run this node with nothing wired to image or mask at all - if a downstream node then complains about missing data, check whether one of the two disable switches is on, or whether you simply forgot to connect the input in the first place; the schema doesn't distinguish those two cases for you.

Since it works on both an image and a mask together, keep an eye on whether the two actually match in size and batch count coming in - this node doesn't claim to reconcile mismatches (that's what flow_tensor_Unify, elsewhere in this pack, is for), it just forwards or blocks what it's given.

And the general caveat for this pack: it's one large Python package covering loaders, samplers, and a long list of utility nodes like this one, so an import error anywhere in it can take this node down too. If it's missing after install, check the ComfyUI console at boot for the actual traceback before re-checking your workflow.