Queue Trigger

Normally you hit "Queue Prompt" and the graph runs once. Queue Trigger lets the graph queue itself again when it executes, which is how you build workflows that repeat without you sitting there clicking. It's the seed of Impact Pack's experimental loop machinery - pair it with the pack's logic nodes and you can run a workflow N times, or keep going until some condition flips. Its category literally says Logic/_for_test, so set expectations accordingly: this is experimental tooling, not a polished feature, and it behaves in ways that surprise you if you don't respect its rules.

How it works

When the node runs, it adds a new prompt to the queue - effectively telling ComfyUI "run this whole thing again." The important subtlety is that it only fires when the signal's status changes, so it won't blindly re-queue forever on its own; something upstream has to change the signal to trigger the next round. That's the mechanism you exploit to make loops that terminate: drive the signal from a counter or a condition, and when it stops changing, the re-queueing stops.

Because it's a pass-through, you thread it into your graph on the signal line so it executes as part of the run rather than as a dead-end.

The inputs and outputs that matter

• signal (any type, *) - the wire it sits on. Whatever you pass in comes back out unchanged; the node's real effect is the side effect of queueing. It re-triggers when this signal's status changes.
• mode (BOOLEAN, default on) - the on/off switch for whether the trigger is armed. Turn it off to run the graph without the auto-queue kicking in.
• Output: signal_opt (the same * value passed through) so you can continue the chain downstream.

It's also an output node, meaning it's a valid endpoint that forces its branch to execute.

How to install it

ComfyUI Manager: search ComfyUI Impact Pack, Install, restart.

Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/ltdrdata/ComfyUI-Impact-Pack
cd ComfyUI-Impact-Pack
pip install -r requirements.txt

Run the pip in ComfyUI's Python environment and restart. No models.

Common issues & troubleshooting

It queued once and stopped, or it never stops. Both come from the "only fires when the signal changes" rule. If nothing upstream is changing the signal between runs, it won't re-trigger; if something changes it every single run, it'll loop indefinitely. The loop's termination lives in your logic controlling that signal, not in this node.

Sequencing matters - put loop controls at the end. Impact Pack's loop and control nodes (Queue Trigger, Control Bridge, Set Widget Value, Set Mute State) are order-sensitive. The pack's guidance is to structure your workflow so these run last, after the real work, or you get out-of-order state changes and misfires. Don't scatter them mid-graph.

Runaway queues. If you kick off an unterminated loop, ComfyUI will keep churning. Clear the queue from the UI to stop it, then fix your termination condition before running again.

It's experimental - treat it that way. The _for_test category isn't decoration. There's no formal tutorial for the loop nodes yet; the pack ships an example loop workflow, and building from that is more reliable than assembling one blind.