CLIPVisionEncode_motorway_edition

CLIPVisionEncode is how you turn an image into conditioning - the visual twin of CLIPTextEncode. It runs a reference picture through a CLIP-Vision model and spits out a CLIP_VISION_OUTPUT that IP-Adapter and similar nodes use to inject style or subject from a reference image. This _motorway_edition clone does the identical thing, except both the input model and the output ride agilly1989's keyed "motorway" bag instead of dedicated wires.

If you've never wired one of these: you feed it a clip_vision (loaded by a CLIPVisionLoader) and an image, pick a crop mode, and get a CLIP_VISION_OUTPUT that downstream nodes read. That's the whole job. The motorway version just relocates the clip_vision and the image into the bag and reads them back by key.

The fields, all six of them

The schema is small and fully keyed except for one dropdown:

• MOTORWAY 🚌💨 - the bag, in and out.
• INPUT_clip_vision_key (default clip_vision) - bag key for the loaded CLIP-Vision model. So upstream you'd have a CLIPVisionLoader (or its _motorway_edition twin) that put the model in the bag under exactly that key.
• INPUT_image_key (default image) - bag key for the reference image.
• crop - an enum, center or none. The one widget you'll actually touch. center center-crops the image before encoding (good for face/subject crops), none encodes the whole frame.
• OUTPUT_CLIP_VISION_OUTPUT_key (default CLIP_VISION_OUTPUT) - where the encoded output lands so an IP-Adapter node can pull it out.

Single output: the motorway itself, which is the family's whole gimmick - one bus wire instead of three.

Install and the reality check

Same pack as every _motorway_edition node, and install is genuinely easy - pure Python, no pip requirements, no model files to fetch:

cd ComfyUI/custom_nodes
git clone https://github.com/agilly1989/ComfyUI_agilly1989_motorway

Restart ComfyUI, or search ComfyUI_agilly1989_motorway in ComfyUI Manager. You'll still need a CLIP-Vision model in your models/clip_vision folder (or equivalent), same as the stock node - the pack doesn't ship models.

The honest warning, same as every node in this pack: it's a single-author beta project, the README banner is "BIG BROKEN WITH ASYNC WILL FIX WHEN I GET THE TIME/MOTIVATION," and the machine-generated clones have been disabled in current releases - __init__.py no longer registers them. Fresh installs give you MotorwayStart and the Motorway AxB ramps; whether this clone shows up depends on which version you land on. If it's missing, the stock CLIPVisionEncode plus a couple of ramps does the same job with more wires and fewer moving parts. And the classic pipe error - 'MotorwayClass' object has no attribute 'hash_' - means one of your key strings doesn't match anything that was written into the bag. Check the keys before you suspect the node.