AC_Super_Controlnet

AC_Super_Controlnet is this pack's single-node version of the core ControlNetApply chain. It takes your positive and negative conditioning, a ControlNet model, and a hint image, and returns the two conditionings with the ControlNet's influence baked in. One node, one wire in and out, no intermediate apply step.

How it works

Under the hood it's the standard ComfyUI machinery: load_controlnet on the file you pick, then set_cond_hint with your strength and guidance window, applied to both conditionings. Notably it sets control_apply_to_uncond to False - the negative conditioning isn't pushed into the unconditional path, matching how the stock ControlNet apply node behaves. And if you set strength to 0, it just passes your conditionings straight through untouched. Clean.

The inputs that matter

• positive / negative - conditioning, usually from AC_Super_CLIPEN or a core CLIP Text Encode.
• control_net - a dropdown of every .safetensors in your models/controlnet folder.
• image - an image upload dropdown. There's also an optional source_image IMAGE input if you want to wire a preprocessed tensor from elsewhere in the graph instead of picking a file.
• strength (0–10, default 1.0) - how hard the condition pulls. From the ControlNet playbook: 0.3–0.7 for loose guidance, 0.8–1.2 for strict adherence to the hint.
• start_percent / end_percent (0–1) - the guidance window. Releasing the condition partway through denoising (end_percent around 0.5) is the classic trick for letting the model add its own detail after composition locks in.

The outputs

positive and negative, both CONDITIONING, ready to go straight into a sampler. That's it - the ControlNet isn't returned, because it's already inside the conditioning.

Installing it

Ships in ComfyUI_AC_FUNV8Beta1 (author Cc啊程, Bilibili ComfyUI teacher; README is Chinese). ComfyUI Manager, search "AC_FUN", or:

cd ComfyUI/custom_nodes
git clone https://github.com/A719689614/ComfyUI_AC_FUNV8Beta1

Restart ComfyUI. No ControlNet files are bundled - you need your own .safetensors in models/controlnet, and it must match your base model family (an SD 1.5 ControlNet won't work on SDXL or Flux).

Where people get burned

Two classic traps. First, this node does no preprocessing - it expects a ready hint image (canny edges, depth map, pose skeleton). Feed it a plain photo and you'll get noise-shaped output. Second, the image dropdown defaults to the pack's bundled example.png, which is a blank white square. If you don't pick your own image, you're conditioning on nothing and wondering why nothing happens. Pick the image or wire source_image, then tune strength and the window.