Upscale Model Loader

This loads an upscale model - an ESRGAN-family .pth like 4x-UltraSharp or RealESRGAN - off disk so an upscaling node can use it. Like the other WAS loaders, its one value-add over core ComfyUI's version is that it also outputs the model's name as a string, which is handy for stamping filenames or logging which upscaler ran. Otherwise it does exactly what core's "Load Upscale Model" does.

To be clear about scope: this loads a pixel upscaler, the non-generative kind. It adds resolution, not detail. The upscaling KB draws the line cleanly - ESRGAN models "add no new content. Fast, predictable, no hallucination risk," appropriate "when the source image already has all the detail you want - you just need more pixels." If you need a model to invent detail on a soft or damaged image, that's a different tool (SeedVR2, SUPIR) and a different loader.

How it works

You pick a .pth upscale model from your ComfyUI/models/upscale_models/ folder and the node loads it into an UPSCALE_MODEL object. That output goes into an "Upscale Image (using Model)" node, which is what actually runs your image through the network. The loader doesn't upscale anything itself - it's the load half of a two-node pair.

The model you pick is the whole ballgame here. From the KB's shortlist: 4x-UltraSharp is the community default for general photographic upscales (clean, not over-sharpened), 4x_NMKD-Siax and the anime-specific RealESRGAN/AnimeSharp variants are the picks for illustration. Different models, wildly different looks, same node.

Inputs and outputs that matter

The input that matters is the model name dropdown, populated from models/upscale_models/. The outputs are the UPSCALE_MODEL (into your upscale-with-model node) and the model name string (into a save filename or log).

How to install it

ComfyUI Manager: search was-node-suite-comfyui, install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/WASasquatch/was-node-suite-comfyui

then install the pack's requirements.txt in your ComfyUI venv and restart. The loader needs no dependencies of its own - but it needs an actual upscale model on disk. Download an ESRGAN .pth (4x-UltraSharp is the safe first pick) into ComfyUI/models/upscale_models/ and restart so the dropdown sees it.

Common issues & troubleshooting

The dropdown is empty. No models in ComfyUI/models/upscale_models/. Drop a .pth in there and restart or refresh. These don't ship with ComfyUI - you download them.

The upscale looks over-sharpened or crunchy. That's the model, not the loader. Some ESRGAN variants are aggressive; try 4x-UltraSharp for a cleaner result, or an anime-tuned model if you're upscaling illustration. And remember a pixel upscaler can't fix a soft source - it'll just enlarge the softness. For that you want a generative upscaler.

Why not just use core's loader? Only reason is the name-string output. If you don't need it, core's Load Upscale Model is one fewer dependency, and this pack is a heavy one to carry for a single string. Use the WAS version when you're already in the suite and want your saves labeled with the upscaler.

Whole pack won't import after a ComfyUI update. The recurring WAS Node Suite issue - it downgrades some packages for older modules, so a ComfyUI update can break the import. Re-run the pack's requirements against the activated venv, or install.bat. Maintenance-only since late 2023.