RM Styles Pipe

The full RM Styles loader spits out eight outputs: positive, negative, motion, three kinds of LoRA lists, and the prompt's number and name. That's eight strands of spaghetti across your canvas every time you want a prompt library in a workflow. RM Styles Pipe is the same engine with a different deal: it packs all of that into a single RM_STYLES_PIPE output, so the graph stays tidy. One wire in, and you unpack it later with the pack's RM Styles Pipe Out node.

It's part of ComfyUI-RMAutomation by Moser9815 - the same pack as RM Styles Full, so it shares the parent class and every mode behaves identically: Manual, Random (with an anti-repeat history of 30), and Increment/Decrement that wrap between your minimum and maximum bounds. It reads the same data/styles.json (or a custom file), caches by modification time, and snaps to the nearest available style number if you ask for one that isn't there. If you've read the Full article, none of the mechanism is new - this is purely about the plumbing.

What's different from the plain loader

Two things, both worth knowing before you reach for it. First, there's no use_prefix toggle on this node - the inputs are just mode, previous_prompt, next_prompt, minimum, maximum, and custom_json_path. The prefix logic was moved downstream to the Pipe Out node, where the Pony quality-tag toggle actually does something. Second, the node is flagged as an output node (OUTPUT_NODE = True) so its frontend callbacks fire - which also means ComfyUI treats it as a valid execution endpoint and the pack's JS hooks run to advance your increment/decrement counters. It also fires the pack's workflow_start progress event and clears stale preview images when a run begins, so if you're using the progress-reporting bits of RMAutomation, this is where they light up.

The output

Just one: Styles Pipe (RM_STYLES_PIPE). Inside the tuple rides the positive and negative prompts, the video motion prompt, the image LoRAs and the high/low motion LoRA strings (already formatted <lora:path:weight>), plus prompt number and name. Feed it straight into RMStylesPipeOut - nothing else in the pack unpacks it, though any node that declares RM_STYLES_PIPE as its input type will accept it.

Why you'd reach for it

Honestly, it's for when the canvas starts to look like a plate of noodles. If you're chaining several style sources into one batch pipeline, or you keep your positive/negative wiring in one place and want the style loader to stay a compact box, the pipe keeps the graph readable without changing behavior. The one trade: the pipe is opaque - no prompt text displayed on the node, no "which style am I on" at a glance. If you need visibility, use the Display variant of RM Styles Full, or drop a text preview node after the Pipe Out.

Installing

Same pack, same steps. ComfyUI Manager → search "RMAutomation" → install, or:

cd ComfyUI/custom_nodes
git clone https://github.com/Moser9815/ComfyUI-RMAutomation

Restart ComfyUI. No Python dependencies, no model files - the whole pack is plain Python plus a bit of JS, so there's nothing extra to babysit.

One caveat

The bundled styles.json is the author's own, pre-filled library - replace it or set custom_json_path before you rely on this in a real workflow. And remember the pipe carries no preview: if your output suddenly looks wrong, check the number it's actually loading in the console log, not the node title.