BasicPipe -> DetailerPipe

This is a plumbing node, and once the plumbing clicks the whole Impact Pack detailing world gets a lot less intimidating. A BASIC_PIPE is Impact Pack's way of shoving model, clip, vae, positive and negative conditioning into one fat cable so you're not dragging five wires across the canvas. A DETAILER_PIPE is that same bundle plus the detection gear a detailer needs - a bbox detector, an optional SAM model, a wildcard. BasicPipe -> DetailerPipe is the adapter that upgrades the first into the second.

You reach for it when you've already got a tidy basic pipe feeding your samplers and now you want to run FaceDetailer (pipe) or Detailer (SEGS) off the same setup without rebuilding all the model connections by hand.

How it works

There's no magic here - it takes the five things inside your basic pipe and repackages them alongside the detector and wildcard you attach, emitting a single DETAILER_PIPE. That pipe then drives the pack's pipe-flavored detailer nodes, which pull the model, prompts, and detectors back out as needed. The point is reuse: build your model bundle once, and every detailer stage taps the same cable instead of you re-wiring the checkpoint into each one.

The inputs that matter

• basic_pipe - your existing model/clip/vae/positive/negative bundle. Required.
• bbox_detector - the detector that finds regions to fix. Required, and the whole reason you're upgrading the pipe.
• wildcard - a text box for per-detection prompt overrides using Impact Pack's __wildcard__ and {a|b|c} syntax. Leave it empty and the detailer just uses the pipe's positive prompt.

The two "Select to add..." dropdowns are UI helpers that paste a chosen LoRA or wildcard token into that wildcard box - they don't add inputs, they're convenience pickers. The optional slots are where the node earns its keep: sam_model_opt for a SAM model to tighten masks to the actual silhouette, segm_detector_opt for a segmentation detector, and detailer_hook for advanced behaviors like noise injection. Output is one detailer_pipe that goes straight into a detailer.

How to install it

Easiest path is ComfyUI Manager: search ComfyUI Impact Pack, click install, restart. By hand: cd ComfyUI/custom_nodes && git clone https://github.com/ltdrdata/ComfyUI-Impact-Pack, install requirements into ComfyUI's Python (pip install -r requirements.txt, or the python_embeded\python.exe -m pip form on Windows portable), and restart. Auto-install was dropped back in v7.6, so the requirements step isn't optional. A small SAM model downloads into ComfyUI/models/sams on first run.

Common issues

The recurring trap is the bbox_detector slot with nothing to plug into it. The node that provides a real YOLO detector, UltralyticsDetectorProvider, was moved out of the base pack in v8.0 and now ships in the separate ComfyUI-Impact-Subpack. Install that, and your detector provider appears. Until you do, you can't populate this pipe with a working detector.

Type-mismatch warnings when you connect the pipe are usually noise, not errors - Impact Pack uses wildcard types for a lot of these connections, and the README says as much: type validation "may still produce error messages" even when things work. If a detailer downstream genuinely refuses the pipe, check that the detailer node is the pipe variant (FaceDetailer (pipe), not plain FaceDetailer) - they don't take the same cable.