Nodes/was-node-suite-comfyui/CLIP Vision Input Switch
ComfyUI Node Runs on cloud

CLIP Vision Input Switch

One checkbox instead of two workflows

By WASasquatch·Created 3 years ago·Updated about a year ago· 1,812
CLIP Vision Input Switch

      WAS's README is direct about this one: "Switch between two CLIP Vision inputs based on a boolean switch." It's a small node doing a small job, but it's part of a pattern that shows up all over this pack - WAS ships an entire family of input-switch nodes (Model, CLIP, Latent, Image, VAE, ControlNet Model, Upscale Model, LoRA, and this one for CLIP Vision), all built on the same idea: plug in two of something, plug in a boolean, and whichever one the boolean points at is what actually flows downstream.

      The reason this family exists at all is that ComfyUI, historically, doesn't give you much in the way of built-in conditional routing. If you want a workflow that sometimes uses one reference-image encoder and sometimes another - a heavier CLIP Vision model for detail-critical runs, a lighter one when you're iterating fast, say - the naive approach is maintaining two nearly-identical workflow graphs and swapping between them by hand. The switch-node family is WAS's answer: build the workflow once, wire both options into the switch, and flip a single boolean instead of rewiring or duplicating the whole graph.

      That boolean doesn't have to be something you toggle manually, either. Feed it from Text Find checking for a keyword, from a Logic AND/OR chain, or from any other boolean-producing node in the suite, and you've got a workflow that reconfigures itself based on some upstream condition rather than needing a human to flip a switch every run. That's the real payoff of this whole family: composability. A boolean from anywhere in your graph can drive a routing decision anywhere else.

      The shape of the node follows directly from the description: two CLIP_VISION inputs and one BOOLEAN input, with a single CLIP_VISION output carrying whichever of the two the boolean selected. Wire that output into whatever downstream node - an IP-Adapter loader, a vision-conditioned sampler - expects a CLIP Vision model, and the rest of your graph doesn't need to know or care which of the two actually got used.

      Installing it: ComfyUI Manager, search "WAS Node Suite," install, restart - the fast path for most people. Manual install: cd ComfyUI/custom_nodes && git clone https://github.com/WASasquatch/was-node-suite-comfyui, then pip install -r requirements.txt (portable: python_embeded\python.exe -s -m pip install -r requirements.txt; manual/venv: activate the venv first), then restart ComfyUI. No model or special dependency tied to this node specifically - it's pure routing logic, ready as soon as the suite loads.

      Where it bites: the one real trap with any switch node in this family is leaving one of the two inputs unconnected and then having the boolean select it - you'll get an error about a missing input rather than the switch quietly falling back to the connected one, so both sides need something wired in even if you only ever expect to use one of them in practice. Beyond that, the standing caveat for the whole pack: WAS Node Suite has been unmaintained since December 2023 (its README title says "Retired"), and the community's recurring complaint is the entire suite failing to import after a ComfyUI update rather than individual nodes breaking in isolation - check your console for a whole-suite import error before troubleshooting this node specifically.

      Categoryx

      Inputs (0)

      No inputs

      Outputs (0)

      No outputs