CLIPTextEncodePixArtAlpha_motorway_edition
PixArt-α's text encoder, on a keyed bus instead of wires
- MOTORWAY 🚌💨
- MOTORWAY 🚌💨
PixArt-α is the odd middle child of the text-to-image family: it's a transformer-based diffusion model, not a U-Net, but it still conditions through a CLIP-style text encoder rather than a giant LLM. The stock node for that is CLIPTextEncodePixArtAlpha, and this _motorway_edition is its clone for agilly1989's motorway pipe pack. Same encode math, different plumbing: the CLIP model is pulled out of a shared keyed bag, and the resulting conditioning is pushed back into that same bag.
For context: this sits under advanced/conditioning in the pack's menu, which is a hint about who it's for. PixArt checkpoints are rarer than SDXL ones these days, so you're probably here because a workflow you loaded uses this exact model family. If you don't own a PixArt checkpoint, this node is dead weight - the motorway version doesn't change what the original does, and you can't fake the encoder it needs.
What's on the node
The info schema is compact. The MOTORWAY 🚌💨 input is the bag. Then:
INPUT_clip_key(defaultclip) - where the PixArt CLIP lives in the bag.widthandheight- the resolution hints PixArt encodes against; defaults to 1024×1024, which is PixArt's native territory.text- the prompt, multiline. The one field you actually write.OUTPUT_CONDITIONING_key(defaultCONDITIONING) - the bag key the conditioning lands under.
Output is the motorway, as always. Under the hood it instantiates the real CLIPTextEncodePixArtAlpha, calls its function with the values it pulled from the bag, and stores the result by key. The 77-token CLIP boundary still applies - PixArt is not a long-prompt model, so keep the prompt tight.
Install and the honest caveat
Same pack, same trivial install - pure Python, no pip requirements, no model files:
cd ComfyUI/custom_nodes
git clone https://github.com/agilly1989/ComfyUI_agilly1989_motorway
Restart ComfyUI, or grab it from ComfyUI Manager under ComfyUI_agilly1989_motorway.
The reason this article ends with warnings: these _motorway_edition nodes were machine-generated in one pass (v1.1.0) and the author has since pulled the plug on them - the loader that registered them is commented out in __init__.py and the README's top line is "BIG BROKEN WITH ASYNC WILL FIX WHEN I GET THE TIME/MOTIVATION." Fresh installs get the Motorway AxB ramp nodes but may not get this clone. If it's in your menu, it works like any other keyed pipe node; if it's not, don't fight it - the plain CLIPTextEncodePixArtAlpha plus a Motorway 0x1 output ramp reproduces this exactly. And remember the motorway's one real error: 'MotorwayClass' object has no attribute 'hash_' means a key you referenced was never written into the bag. That's the price of hiding your wiring.
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| MOTORWAY 🚌💨 | MOTORWAY 🚌💨 | — | |
| INPUT_clip_key | STRING | clip | — |
| width | INT | 10240–16384 | — |
| height | INT | 10240–16384 | — |
| text | STRING | — | |
| OUTPUT_CONDITIONING_key | STRING | CONDITIONING | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| MOTORWAY 🚌💨 | MOTORWAY 🚌💨 | — |