UpscaleModelLoader_motorway_edition

When an image comes out of the sampler at the wrong size, your first stop is an ESRGAN-style upscaler - RealESRGAN, 4x-UltraSharp, that whole family. They're small standalone model files that live in models/upscale_models, and UpscaleModelLoader is the stock node that loads one so an upscaling node can use it. This _motorway_edition clone does the same load through the agilly1989 Motorway pipe, so instead of a dedicated UPSCALE_MODEL wire you just set a key and the loaded model lands in the shared dict.

Motorway, in one breath: start with MotorwayStart, then each _motorway_edition node trades data wires for short key strings, reading and writing a shared dict that flows down one MOTORWAY connection. These nodes are auto-generated clones of ComfyUI's own classes, so underneath this is the plain UpscaleModelLoader - same file support, same formats.

What it does. Pick model_name from the dropdown of everything in models/upscale_models, and the node loads the upscale model into the Motorway under the key you choose via OUTPUT_UPSCALE_MODEL_key (default UPSCALE_MODEL). That's it. The loader does not upscale anything - it just prepares the model. You still need the upscaling step itself (the stock ImageUpscaleWithModel node, or your preferred alternative) wired into whatever consumes the result. On the Motorway, that consumer pulls the upscale model by key, so the whole "load then upscale" chain becomes two key strings instead of a cable run.

The inputs that matter. MOTORWAY is optional, as it is for all loaders here - connect an existing Motorway to add to it, or leave it unconnected to act as a fresh start. model_name is the only real choice. OUTPUT_UPSCALE_MODEL_key is the key the model is stored under. Single output: the MOTORWAY port.

Common issues. The Motorway's usual traps apply - keys are exact-match and case-sensitive, and 'MotorwayClass' object has no attribute 'hash_' means you referenced a key that was never stored. Worth knowing from the upscaling side too: an ESRGAN model only upscales, it doesn't add detail beyond what it was trained on, and throwing 4x at a low-res latent decode is how people end up with plastic-looking output. Upscale in sensible steps. And the standing caveat: the pack's current main branch has the clone registration commented out in __init__.py, so if the node's missing from your menu, that's the version you're on, not a broken install.

Install via ComfyUI Manager (search "ComfyUI_agilly1989_motorway") or git clone https://github.com/agilly1989/ComfyUI_agilly1989_motorway into custom_nodes, restart. No models to download (the upscaler files themselves go in models/upscale_models as usual), no dependencies. Solo-dev beta pack - treat the Motorway as a convenience layer and you'll be fine.