ToDetailerPipe

The detailer nodes in Impact Pack - FaceDetailer (pipe), Detailer (SEGS/pipe) - need a lot of inputs: model, clip, vae, positive and negative conditioning, plus a detector. Wiring all of those into every detailer individually turns your graph into spaghetti fast. ToDetailerPipe packs them into a single DETAILER_PIPE wire so you route one line instead of six or seven. It's pure convenience, and once you've built a workflow with two or three detailing passes you'll never go back.

A DETAILER_PIPE is a superset of a basic pipe: it carries the sampler ingredients and the detector (and optionally a SAM model and a detailer hook), because that's the full set a detailer needs to detect and refine.

How it works

You plug the individual pieces in once, the node bundles them, and downstream detailer-pipe nodes unpack them internally. Change one ingredient - swap the checkpoint, adjust the prompt - and every detailer fed by that pipe picks it up. If you need to tweak the bundle later without rebuilding it, the pack has an Edit DetailerPipe node for that; ToDetailerPipe is the one that assembles it in the first place.

The inputs and outputs that matter

The required inputs are the whole sampler + detection kit:

• model, clip, vae - your checkpoint's three pieces.
• positive / negative - the conditioning the detailer will sample with.
• bbox_detector - the detector that finds regions to fix. This is what makes it a detailer pipe and not just a basic one.
• wildcard - an optional prompt string applied during detailing (handy for a per-region prompt like "detailed eyes").

Two of the "inputs" - Select to add LoRA and Select to add Wildcard - are UI helper dropdowns that insert LoRA/wildcard syntax into the wildcard text for you, not data connections.

Optional slots let you enrich the bundle: sam_model_opt (SAM for tighter masks), segm_detector_opt (a segmentation detector), and detailer_hook (advanced per-pass behavior).

The single output is detailer_pipe (DETAILER_PIPE), which feeds any detailer-pipe node.

How to install it

Comes with the Impact Pack. ComfyUI Manager: search ComfyUI Impact Pack, Install, restart. Manually:

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

then restart. The bbox_detector you plug in usually comes from the Impact Subpack (YOLO detectors), which since v8.0 installs separately.

Common issues & troubleshooting

Type-mismatch errors when connecting to a detailer. Make sure you're feeding a pipe detailer (the ones with "pipe" in the name). A DETAILER_PIPE won't connect to the non-pipe variants that expect separate model/clip/vae inputs.

The detailer complains about a missing detector. bbox_detector is required here - the pipe carries the detector, so if you left it unconnected, the downstream detailer has nothing to detect with.

You want SDXL with a refiner. This node builds a standard detailer pipe. For the SDXL base+refiner setup, use BasicPipe -> DetailerPipe (SDXL) instead, which takes two basic pipes (base and refiner) and wires them for refiner-assisted detailing.

You need to change one ingredient mid-workflow. Don't rebuild the whole pipe - use Edit DetailerPipe to swap a single element, which keeps the rest of the bundle intact.