Mute By Title

If you've read the Bypass By Title article, this one is 80% the same node wearing a different coat. Same title-matching, same instant canvas behavior, same single switch - but mute instead of bypass, and that one letter difference matters more than it sounds.

Bypass vs mute, the one paragraph you need

Bypass skips a node and passes its inputs straight through, so the graph keeps running - a bypassed KSampler just hands its latent along. Mute actually stops execution dead at that node: the workflow halts there. In practice, muting is what you reach for when you want to cut a branch out - a whole img2img pass, a post-processing step, an experiment - and bypass is what you use when you want a node to sit there inert but still let data flow.

What it does

AUNSetMuteByTitle takes a list of node titles and toggles mute state on every match. The titles field is a multiline string - newline, comma, or semicolon separated - and supports the same ! / - exclusion prefixes as its bypass twin, so refine, !face matches anything titled "refine" but leaves "face" alone. The Switch is a boolean labeled Active 🟢 / Mute 🔇; flip it and the matched nodes mute or wake up on the canvas instantly, no queue run required.

No outputs - it's a controller, not a data node. It sits on the graph, does its thing, and gets out of the way.

How it works

Same architecture as the bypass version. The Python node parses your titles and fires a AUN_set_mute_by_titles WebSocket event; frontend JavaScript matches titles against the graph and flips node modes, recursively for grouped nodes. It's instant, it's visual, and it doesn't touch the queue - which is the entire point of this family of nodes.

When to actually use it

The realistic use case is version-switching inside one workflow. Say you have a high-detail pass that sometimes produces artifacts: leave it in the graph, mute it by title when you don't want it, and the run stops before it - instead of deleting and re-adding nodes or keeping two nearly identical workflows. The same logic applies to LoRA stacks, refiner passes, or any optional branch you toggle between generations.

The caution is the same as its twin: it matches by title, so it only behaves predictably if your nodes have deliberate names. If half your graph is still called "KSampler," a mute-by-title call will take out every KSampler in the workflow. Rename things, and this becomes surgical.

Installing

It's part of the AUN ComfyUI Nodes pack:

cd ComfyUI/custom_nodes
git clone https://github.com/loz2754/AUN-ComfyUI-Nodes

Restart ComfyUI, or use ComfyUI Manager and search "AUN". If you manually cloned and hit missing-module errors, the pack's requirements are in its requirements.txt (piexif, opencv-python-headless, requests, imageio-ffmpeg):

pip install -r custom_nodes/AUN-ComfyUI-Nodes/requirements.txt

Which one should you grab? If you're not sure, you probably want bypass - it's gentler, keeps the graph alive, and is the more forgiving default. Reach for mute when you genuinely want execution to stop at a branch. Between the two, most people end up with one of each in the toolbox.