Nodes/JPS Custom Nodes for ComfyUI/IP Adapter Settings Pipe (JPS)
ComfyUI Node

IP Adapter Settings Pipe (JPS)

Unpack a single-adapter tuning bundle

By JPS-GER·Created 3 years ago·Updated 2 years ago· 100
IP Adapter Settings Pipe (JPS)
  • ip_adapter_settings
  • ipa_weight
  • ipa_wtype
  • ipa_noise
  • ipa_start
  • ipa_stop
  • ipa_crop
  • ipa_zoom
  • ipa_offset_x
  • ipa_offset_y
  • ipa_mask
  • crop_intpol
  • sharpening
  • ipa_model

Unpacks the bundle from IP Adapter Settings (JPS) - weight, weight-type, noise, timing window, crop/zoom/offset framing, mask mode, interpolation, sharpening, and model choice for one IP-Adapter pass - into individual outputs wherever your graph actually needs them.

What comes out of it

One input, ip_adapter_settings - a BASIC_PIPE (JPS's own bundle, a plain tuple, not Impact Pack's object of the same type name). Thirteen outputs:

  • ipa_weight (FLOAT) - overall adapter strength.
  • ipa_wtype - the weight-distribution curve across sampling steps (linear, ease in/out, ease in-out, reverse in-out, weak input/output/middle, strong middle - the newer nine-option list, not the older three-option scheme its Single-Settings sibling uses).
  • ipa_noise (FLOAT) - reference-embedding noise.
  • ipa_start / ipa_stop (FLOAT) - the step window the adapter is active for.
  • ipa_crop, ipa_zoom (FLOAT), ipa_offset_x / ipa_offset_y (INT) - reference-image framing before it's encoded.
  • ipa_mask (INT) - which of six mask modes constrains where the adapter applies.
  • crop_intpol - the resize algorithm used for the framing crop.
  • sharpening (FLOAT) - post-crop sharpen amount.
  • ipa_model (INT) - which SDXL IP-Adapter variant (ViT-H, Plus ViT-H, Plus Face ViT-H).

This node runs no adapter logic itself - it's routing, handing each value to whatever ComfyUI_IPAdapter_plus apply node and image-crop node are actually doing the work in your graph.

For a workflow using multiple reference images at once - up to two IP-Adapter slots plus two Revision slots, per this pack's own naming conventions - you'd typically place one of these Settings/Pipe pairs per adapter you're actually tuning independently, then route the reference images themselves separately through Images Masks MultiPipe (JPS), which is where the actual IMAGE/MASK data travels rather than the numeric tuning values this pipe carries.

Installing it

Ships with the pack. ComfyUI Manager: search JPS Custom Nodes for ComfyUI, install, restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/JPS-GER/ComfyUI_JPS-Nodes.git

Restart afterward, deleting any earlier copy of the pack first. No models of its own - you'll separately need ComfyUI_IPAdapter_plus (search IPAdapter Plus in Manager) for the actual adapter apply node this pipe is built to feed, plus the SDXL IP-Adapter weights.

Where people get tripped up

Without an IP Adapter Settings (JPS) node feeding it, this pipe has nothing to unpack.

Watch the wiring closely here: this pack ships a near-identical IP Adapter Single Settings Pipe (JPS), which takes an input with the exact same name and type (ip_adapter_settings, BASIC_PIPE) but expects values from the other Settings node - the one with the older three-option ipa_wtype list (original/linear/channel penalty) instead of this one's nine. Since BASIC_PIPE carries no schema ComfyUI checks at connection time, crossing the two pairs will connect without complaint and produce a weight-type value that doesn't mean what you think it does downstream. If your adapter's weight-type behavior doesn't match what you selected in the Settings node, confirm you're actually using the matching Pipe node - Settings and Single-Settings are not interchangeable despite looking almost identical.

CategoryJPS Nodes/Pipes

Inputs (1)

NameTypeDefaultDescription
ip_adapter_settingsBASIC_PIPE

Outputs (13)

NameTypeDescription
ipa_weightFLOAT
ipa_wtypelinear,ease in,ease out,ease in-out,reverse in-out,weak input,weak output,weak middle,strong middle
ipa_noiseFLOAT
ipa_startFLOAT
ipa_stopFLOAT
ipa_cropcenter,top,bottom,left,right
ipa_zoomFLOAT
ipa_offset_xINT
ipa_offset_yINT
ipa_maskINT
crop_intpollanczos,nearest,bilinear,bicubic,area,nearest-exact
sharpeningFLOAT
ipa_modelINT