Default Weights 🛂🅐🅒🅝

This is the no-op weights node, and that's not an insult - it's genuinely useful. It produces a neutral, full-strength ControlNet weights object that works with any ControlNet type, no per-block tuning, no architecture-specific layout. When you don't want to reweight anything but you still need a weights object in the graph, this is the clean way to get one.

Why would you need a weights object if you're not customizing the weights? Two reasons. First, its TK_SHORTCUT output gives you a ready-made Timestep Keyframe to plug into an Apply node's timestep_kf input - a tidy way to start a schedule from neutral defaults. Second, it has a cn_extras input, so you can attach extras (Middle Weight Extras, SparseCtrl-specific multipliers) on top of otherwise-default weights without dragging in a full per-block custom node.

The point of "universal"

The other weight nodes in this pack are shaped for specific architectures - SD 1.5 has 13 blocks, Flux has its own count, T2I-Adapter has 4. This one sidesteps all that. It emits a default weights profile the pack applies universally, so it doesn't care whether the ControlNet downstream is SD 1.5, SDXL, Flux, a T2IAdapter, or a SparseCtrl. If you've hit a "weight type not compatible" error from one of the specific nodes and you don't actually need custom weighting, swapping in Default Weights makes the error go away because it fits everything.

Inputs and outputs

There are no required inputs - that's the whole design. The one optional:

• cn_extras - attach extra weight modifiers here (e.g. from Middle Weight Extras or SparseCtrl Weight Extras) to layer onto the default weights.

Outputs:

• CN_WEIGHTS - the neutral weights, into a Timestep Keyframe's cn_weights or the weights_override on Apply Advanced ControlNet.
• TK_SHORTCUT - a Timestep Keyframe carrying these weights, so you can wire straight into a timestep_kf input and skip building a separate keyframe node.

Installing the pack

Through ComfyUI Manager (Install Custom Nodes → search "ComfyUI-Advanced-ControlNet" → restart) or via cd ComfyUI/custom_nodes && git clone https://github.com/Kosinkadink/ComfyUI-Advanced-ControlNet and a restart. It's Kosinkadink's pack, pure ComfyUI, nothing to compile. Reminder: no preprocessors here (use comfyui_controlnet_aux) and no ControlNet models (those go in ComfyUI/models/controlnet).

Common gotchas

There's not much to break here, which is the point - but people do reach for it expecting it to do something. It doesn't change your output on its own; it's a neutral carrier. If you wanted the ControlNet softened or reweighted, you want Scaled Soft Weights or one of the Custom Weights nodes instead.

The one real use case to remember: it's your escape hatch when an architecture-specific weights node throws a compatibility error and you didn't need the custom weights anyway. And when you want to attach cn_extras to a plain, unweighted ControlNet, this is the least-fuss way to introduce them into the graph.