Select Model 20

What it is

The regular Load Checkpoint node makes you commit to one checkpoint per node - swap models and you're either reconnecting a new loader or maintaining a pile of loader nodes with bypasses. Select Model 20 is a bigger version of ComfyRoll's well-known "CR Select Model" pattern (the author's own README credits it directly), scaled from five slots to twenty: preload up to twenty checkpoints into named slots, then flip which one actually loads with a single integer widget.

It's exactly the kind of quality-of-life node that earns a permanent spot in a workflow once you're doing any serious model comparison - running the same prompt across a shortlist of checkpoints, keeping your most-used twenty models one click away instead of rewiring a loader every time you switch.

How it works

Twenty enum dropdowns, each populated from your ComfyUI checkpoints folder, let you preload up to twenty specific files at once - think of it as pinning your favorites. A separate integer widget, select_model, decides which one of those twenty actually gets loaded and sent downstream this run. Loading itself works the same as a standard checkpoint load: you get back the model, CLIP, and VAE components split out the way every sampler and conditioning node expects.

Inputs that matter

• ckpt_name1 through ckpt_name20 – twenty enum pickers, each showing checkpoints available in your setup. Fill in as many or as few as you actually use; unused slots just sit there.
• select_model (INT, 1–20, default 1) – which of the twenty slots is active. This is the single widget you change to swap your whole workflow to a different checkpoint.

Outputs

• MODEL, CLIP, VAE – the standard three-way split every Load Checkpoint node produces, ready to wire into your sampler and text encoders exactly as usual.
• ckpt_name – a STRING echoing back which file got selected, useful for stamping into a saved filename so you can tell which checkpoint produced which output later.

Installing it

Comes with the rest of the pack:

cd ComfyUI/custom_nodes
git clone https://github.com/Shadetail/ComfyUI_Eagleshadow

Or search "Eagleshadow Custom Nodes" in ComfyUI Manager. No extra dependencies - it loads whatever checkpoints you already have.

Common issues

Twenty slots means twenty checkpoints loaded into your VRAM/RAM budget conversation - sort of. The node only actually loads the one you select each run; the other nineteen dropdowns are just picking a filename, not pre-loading the weights. Don't worry that filling in all twenty slots means twenty models sitting in memory at once - it doesn't work that way.

Slots show "None" until you've got checkpoints installed. The dropdowns are populated live from your checkpoints folder. If a slot looks empty, that's a model you haven't downloaded yet, not a bug in the node.

Every checkpoint you preload needs matching CLIP/VAE expectations downstream. This node outputs a single MODEL/CLIP/VAE set per run based on whichever slot is selected - if you're switching between architectures that need very different conditioning setups (say, an SDXL checkpoint in one slot and a Flux checkpoint in another), make sure the rest of your graph downstream can actually handle whichever one gets selected, since the node itself won't adapt your text encoding or sampler settings for you.

Keep it for your actual favorites, not everything you own. Twenty slots sounds like a lot until you have more checkpoints than that - treat this as a curated shortlist node, not a full model library browser. If you need to browse your entire collection, a different node (or the standard loader with manual swapping) is the better fit.