Load Checkpoint+
A checkpoint dropdown that doesn't make you scroll 300 files
- MODEL
- CLIP
- VAE
- ckpt_name
Load Checkpoint+ is what you get when you've accumulated a wall of checkpoints and the built-in loader's flat dropdown has become a search-free slog. It adds a text filter that narrows the list as you type, groups the results with section headers, and hides the .safetensors/.ckpt/.pt noise so the list reads like model names instead of filenames.
It's part of ComfyUI-FBnodes, François Beaudry's utility pack - one of the Load Model+ family that also covers diffusion models and LoRAs. Same author as ComfyUI-Prompt-Manager, and it shows: the whole point of this node family is that you're supposed to find the model you loaded weeks ago, not scroll past it.
How it works
Two inputs sit on the node: filter and ckpt_name. Type comma- or space-separated terms into filter - say sdxl, anime - and the dropdown only shows checkpoints whose name or path matches all of them. It's an AND match, so you can stack "sd" and "xl" and "realistic" and drill straight to a specific file. The matching runs against the full path, which is a quiet lifesaver: it means anime finds files in anime/ subfolders even when the filename says nothing about anime.
The frontend also groups the results with section headers (the README shows the style as ----Anime----) and, critically, treats the headers as safe to click - selecting a section header auto-selects the first real model in it, so you can't accidentally "load" a header. The display strips extensions so you see juggernautXL_v9 rather than juggernautXL_v9.safetensors.
Underneath it's a normal checkpoint loader: it calls ComfyUI's standard load_checkpoint_guess_config, resolves the display label back to the real path, and outputs the three things you expect - MODEL, CLIP, VAE - plus a ckpt_name COMBO output so downstream nodes (like a filename saver) know which one you picked.
The inputs
filter(STRING) - the search box. Leave empty to see everything.ckpt_name(COMBO) - the filtered dropdown; also a place you can type or paste a path.
Outputs: MODEL, CLIP, VAE, and ckpt_name (COMBO).
Installing it
ComfyUI Manager - search "ComfyUI-FBnodes" - or:
cd ComfyUI/custom_nodes
git clone https://github.com/FranckyB/ComfyUI-FBnodes.git
pip install -r ComfyUI-FBnodes/requirements.txt
Restart ComfyUI. The only real dependencies (av, color-matcher) belong to the pack's video nodes; this node itself is pure ComfyUI and needs no model files beyond whatever checkpoints you already have.
Gotchas
The filter is per-session UI state, so a filter you left in place hides everything else next time you open the workflow - clear it when you're hunting for a different family. And remember the ckpt_name output is a display-friendly label, not necessarily the path ComfyUI internals use; the node resolves the real path internally, so don't try to feed that label back into a plain loader expecting a raw folder-relative path. Finally, if a checkpoint sits in a deep subfolder, its path is part of the match target - that's a feature, but it means wan can match a lot of unrelated files that happen to live in a wan/ folder.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| filter | STRING | Filter checkpoints by multiple terms in name/path (AND match; use spaces/commas, e.g., 'sdxl anime') | |
| ckpt_name | COMBO | Select checkpoint (filtered by search text above) |
Outputs (4)
| Name | Type | Description |
|---|---|---|
| MODEL | MODEL | — |
| CLIP | CLIP | — |
| VAE | VAE | — |
| ckpt_name | COMBO | — |