abyz22_Convertpipe

ComfyUI wires get messy fast - model, clip, vae, both conditionings, an image, a latent all crossing the canvas is five cables you have to keep straight. Pipes are the standard answer: bundle them into one object and drag a single wire. This pack's abyz22_Topipe bundles its own flavor of pipe (model, clip, vae, positive, negative, image, plus two latents), and abyz22_Convertpipe is the adapter that translates that bundle into the format Impact Pack's nodes actually consume.

It lives in abyz22/image_control, a fork of ltdrdata's Impact Pack, so the two pipe dialects coexisting in one repo is the backstory. Impact Pack speaks BASIC_PIPE - a plain 5-tuple of (model, clip, vae, positive, negative). This pack's pipe is a dict with extra slots. Convertpipe does the translation.

What it does

One input, two outputs:

• pipe - the extended pipe from abyz22_Topipe.
• basic_pipe - the same model/clip/vae/positive/negative repacked as Impact Pack's BASIC_PIPE tuple. Feed this to Impact nodes that take basic_pipe inputs.
• image - the IMAGE slot pulled out alongside it, so you don't lose it in the conversion.

That's the whole surface area. If you're already inside the Impact Pack ecosystem, this is the bridge that lets this pack's pipe flow into Impact's detailers and detectors without rewiring everything.

How it works

The code pulls MODEL, CLIP, VAE, POSITIVE, and NEGATIVE out of the pipe dict and assembles a 5-tuple, then returns that plus the IMAGE slot. It's a straight repack - no model math, no copying, just a dictionary read and a tuple write. The one structural assumption: those five keys must exist in the pipe. abyz22_Topipe always writes them (they're required there), so as long as you built the pipe with Topipe you're safe. If you assembled the dict some other way and it's missing a key, this node throws a KeyError instead of a helpful message.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/abyz22/image_control

or ComfyUI Manager → "image_control", restart ComfyUI. The pack's requirements add openpyxl, pytorch-lightning, kornia.

The catch

The basic_pipe output is only meaningful if Impact Pack is actually installed and you're wiring into its nodes - the type is just a Python tuple, so ComfyUI will happily let you drag it anywhere, and the mismatch only shows up as a runtime failure in the consumer node. If your workflow never touches Impact Pack, Convertpipe is solving a problem you don't have; use abyz22_Frompipe to unpack this pack's pipe directly instead. If you do use Impact, this is exactly the adapter you want, and it's the cleanest way to keep a pipe-based graph tidy across both packs.