RM Styles Pipe Out

RM Styles Pipe Out is the unpacking half of the pack's style-pipe pair: it takes one RM_STYLES_PIPE input and splits it back into the same eight outputs the plain RM Styles Full loader produces. You don't normally start a workflow with this node - you end a tidy pipe run with it. The interesting bit is the single optional input sitting on it: Toggle Pony Tags.

That toggle exists because of how Pony-family checkpoints work. Models in the Pony/Illustrious lineage were trained with a numeric quality-rating system, so you're supposed to prepend tags like score_9, score_8_up, score_7_up to the positive prompt and drop score_6, source_pony, low quality into the negative. If your style library was written before you moved to a Pony checkpoint, every style you load is running without the tags that model was trained to care about. Flip the toggle (tooltip: "Add PDXL (Pony) quality tags to positive/negative prompts") and the node prepends the correct score tags to both sides automatically - a one-click retrofit for an entire library, which is genuinely handy.

How it works

The node takes the pipe tuple and unpacks it into Positive, Negative, Motion Prompt, Image Loras, Motion Loras High, Motion Loras Low, Prompt Number, and Prompt Name - the exact same output names as the loader, so you can swap the loader/pipe-out combo in and out of a workflow without rewiring anything downstream. If the pipe is empty or missing, it returns empty strings instead of crashing, so a broken upstream style source won't take down your CLIP encode. It's a plain pass-through node - not marked as an output node - so it won't terminate the graph; whatever you hang off its outputs keeps executing normally.

The input that matters

Just the one you already have: styles_pipe, wired from RMStylesPipe (or anything else emitting RM_STYLES_PIPE). Toggle Pony Tags is off by default, and it's all-or-nothing - it applies to the whole style, not per-entry. If some of your styles already carry their own score tags, you'll get them twice; the fix is to strip the tags from your JSON and let this node own them.

Also worth a word of caution from how prompt engineering has shifted: those score tags are a CLIP-era, SDXL-lineage convention. On a Pony or Illustrious checkpoint they matter a lot; on 2026 LLM-encoded models the tags are inert, and on those the toggle is just harmless noise. Know your base model before you lean on it.

Installing

Part of ComfyUI-RMAutomation by Moser9815. ComfyUI Manager → search "RMAutomation" → install, or:

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

Then restart. No extra Python packages, no model downloads - like the rest of this pack, it's refreshingly dependency-free.

The honest summary

If you're already using RM Styles Pipe to keep your canvas tidy, this is the node you pair it with, full stop. If you're on a Pony-family checkpoint with a legacy prompt library, the toggle is the reason to add the pair. Otherwise it's a straightforward unpackage - the least glamorous node in the pack, but the one that makes the tidy-wiring trick actually usable.