SB Load Model (SuperBeasts.AI)
The one-job loader that fetches SuperBeasts weights so you don't have to
- model
SB Load Model is a loader with exactly one job: produce an SBMODEL for the Super Pop Color Adjustment node. That's it. It won't load your checkpoints, your LoRAs, or anything else - it exists so the SuperBeasts color model has a home, a version picker, and an automatic download path. If you're here, it's because a Super Pop workflow told you a node was missing, and this is that node.
The good news is that it's one of the most convenient loaders in the ecosystem to set up, because it does the boring parts for you.
How it works
The node reads from a small built-in registry of published SuperBeasts models. Today that's one family with two versions:
SuperPopColorAdjustment/v1.0- the legacy ONNX build (SuperBeasts_ColorAdjustment_512px_V1.onnx). Works on CPU but slowly; you'd only pick this on purpose.SuperPopColorAdjustment/v2.0- the current safetensors build (SuperBeasts_ColorAdjustment_512px_V2.safetensors), which runs natively in standard ComfyUI and is whatlatestresolves to.
Pick model_key = SuperPopColorAdjustment/latest and move on. On first run the node downloads the weights from the official model repo into custom_nodes/ComfyUI-SuperBeasts/models/, verifies the SHA-256 checksum, and prints a one-time license banner to your console. If the file is ever corrupt or tampered with, you get a checksum-mismatch error telling you to delete it and rerun - no silent corruption.
There's also a small model cache (LRU, max 4 entries by default, tunable with the SUPERBEASTS_MODEL_CACHE_MAX environment variable) so repeated runs don't reload and re-allocate the weights every time. If you're on the GPU, set device to GPU (or leave AUTO and let it figure it out); CPU exists for people running inference on the box.
The inputs that matter
Honestly, just the two required widgets:
- model_key - the dropdown of family/version combos.
SuperPopColorAdjustment/latestis the right answer for essentially everyone. - device -
AUTO/CPU/GPU. Leave it on AUTO unless you have a reason not to.
Output is one thing: model, of type SBMODEL, which plugs into the model input of Super Pop Color Adjustment. It's a custom type that only this pack's nodes understand, so there's no ambiguity about where it goes.
Install
It ships inside the SuperBeasts pack, so install the pack and this node comes with it:
cd ComfyUI/custom_nodes
git clone https://github.com/SuperBeastsAI/ComfyUI-SuperBeasts
Restart ComfyUI. Or use ComfyUI Manager - search "SuperBeasts" and install the pack from the registry. No pip installs needed; the pack's dependencies are just torch/Pillow/NumPy, all already present in ComfyUI.
Common issues
- First run downloads a couple of GB. The v2.0 safetensors model isn't tiny, and the download hits GitHub raw. If it fails partway, the node handles it gracefully and you can just rerun - the checksum check will catch a partial file and tell you to delete it.
- You see a big license banner in the logs. That's the SPCA-Community-NoSaaS notice, printed once per install. Local, personal and client use is fine; public SaaS/API redistribution of the weights needs a license from the author. If you're automating,
SUPERBEASTS_SILENCE_LICENSE=1suppresses the banner. - "Unknown model/version combination." You typed a key that isn't in the registry. Use the dropdown, don't type - the dropdown values are guaranteed-valid and keep saved workflows loadable.
One honest caveat: because this is a pack-internal loader, it's dead weight in any workflow that isn't using Super Pop. Don't install the pack for this node - install it for the color adjustment, and think of SB Load Model as the required plumbing.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| model_key | COMBO | 3 options: SuperPopColorAdjustment/latest, SuperPopColorAdjustment/v1.0, SuperPopColorAdjustment/v2.0 | |
| device | COMBO | AUTO | 3 options: AUTO, CPU, GPU |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| model | SBMODEL | — |