Nourivex Upscale Model Loader
The upscale-model dropdown the pack wants you to use instead of the core one
- UPSCALE_MODEL
This is the most transparent node in the pack: a dropdown, one output, and nothing else. Nourivex Upscale Model Loader wraps ComfyUI's core UpscaleModelLoader, so it lists every model in your models/upscale_models folder and hands you the selected one as a ready-to-use UPSCALE_MODEL. There is no hidden logic, no defaulting trickery - it's the same loader wearing a Nourivex label so it matches the rest of the pack's I/O.
What you set, what you get
One input, model_name, which is just the dropdown of upscale weights you've got on disk. One output, UPSCALE_MODEL, which plugs straight into Nourivex Image Upscale With Model - or into the core ImageUpscaleWithModel, since the type is identical and the two worlds interoperate freely.
The pair of Nourivex nodes is meant to be used together: this one loads, the other one applies. In the pack's default workflow the chain is decode → upscale with 4x-UltraSharpV2 (the workflow's go-to weight) → preview, with the route toggle deciding whether the upscale actually happens.
The one thing people miss
Nothing gets downloaded for you. The dropdown only shows what's already in models/upscale_models, and if that folder is empty, the dropdown is empty. The README's workflows assume you've fetched an ESRGAN-family model like 4x-UltraSharpV2 yourself (it's the community standard for the "more pixels, keep it sharp" job - quick, deterministic, adds no hallucinated detail). Drop any .safetensors ESRGAN weight into that folder, restart ComfyUI, and it appears in the list. If your dropdown is empty and you're staring at "where's the model", that's a file problem, not a node problem.
It's also worth knowing which job this is for. The models it loads are pixel upscalers - they make the image bigger without inventing content. If your goal is restoring a soft or damaged image, you want a generative upscaler and this is the wrong tool; if your goal is a clean 4x of a good render, this is exactly the right one. The pack frames it as the final "polish" step, and that's the honest use.
Install
Standard pack install:
cd ComfyUI/custom_nodes
git clone https://github.com/Nourivex/Nour_Comfy
Restart ComfyUI (or search Nour_Comfy in ComfyUI Manager). No Python dependencies. And an honest closing note: on its own this node is strictly redundant with the core UpscaleModelLoader you already have. Its value is the pack's consistency story - all-Nourivex I/O in the shipped workflows, one pack to install, no missing-node surprises. If you're not running the pack, you're not missing anything by sticking with the core loader.
Inputs (1)
| Name | Type | Default | Description |
|---|---|---|---|
| model_name | COMBO | 0 options: |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| UPSCALE_MODEL | UPSCALE_MODEL | — |