IPAdapter Mad Assistant V6

Let's start with the conclusion, because V6 is the version the README won't help you understand: V6 is a copy of V5 with a worse UI. The twelve layer Weight sliders run the exact same -5.0 to 5.0 range, the output is the same comma-joined layer:weight string, the Random mode logic is byte-for-byte identical. The only difference in the shipped code is that V5's sliders carry a "display": "slider" hint and V6's don't - so on V6 the layer weights render as plain number fields instead of draggable sliders.

Why does that matter? It doesn't, functionally. The node still builds the layer_weights STRING output exactly the same way, and it still wires into the IPAdapter Mad Scientist node's layer_weights input (right-click → Convert to Input first, then drag the wire). The difference is purely cosmetic - V6 just looks less like what the pack's README screenshots advertise.

That makes V6 the easiest recommendation in the entire pack: don't use it. Use V5. Same range, same behavior, and you actually get sliders to drag. The only reason V6 exists is presumably history - the V-numbers were added one commit at a time (the last one, "add V6", landed in August 2024), and by the end the author was shipping near-duplicates rather than meaningful steps. This is that pattern made visible.

Everything you need to know carries over from V5. Random mode fills currently-zero layers from the full -5..5 range, which at this spread is chaos - pin non-zero layers if you run it, and probably don't run it here at all. The seed input is declared in the schema but never read by the code, so random mode is unreproducible no matter what seed you enter. For normal style transfer you're better off on the base node (0..1) or V1 (-1..1) anyway; the wide-range nodes exist for the rare "this one layer needs everything" case.

Install

Same single-file pack:

cd ComfyUI/custom_nodes
git clone https://github.com/11dogzi/ComfUI-EGAdapterMadAssistant

Restart ComfyUI, or search ComfUI-EGAdapterMadAssistant in ComfyUI Manager. No pip, no dependencies, no model downloads - pure standard library. You still need ComfyUI_IPAdapter_plus and its model files for Mad Scientist to exist in your graph.

And a warning that applies to the whole pack: the README says "0–10 layers" and calls V2 the "-1 to 1" node. Both are wrong - the code ships 0–11 layers and V1 is the -1..1 one. When the README and the code disagree, the code is the ground truth. In this pack, that happens more than once.