🐸 Load: Model + CLIP + VAE + Name

Frog Load: Model + CLIP + VAE + Name is the plain Frog Loader plus one output that unlocks the pack's automation: a model_name string carrying the diffusion model's filename stem. Pick ANIMA_beta57_v3.safetensors and you get ANIMA_beta57_v3 out the wire. That one string is what the whole "prompt follows the checkpoint" system is built on - it's the glue between a loader and the Library.

Everything else is identical to 🐸 Load: Model + CLIP + VAE. Same three dropdowns for diffusion_model, clip_encoder, and vae_model, same optional Turbo_Lora slot with its lora_strength, same MODEL, CLIP, and VAE outputs. If you only need a combo loader, use the plain one. If you're building the auto-loading setup - where every checkpoint you select drags its matching quality prompt along with it - this is the version you want, because it's the only one that tells you the name of what it loaded.

How the auto-load wiring goes

The intended chain is short and worth seeing in full:

1. This node loads your model and outputs model_name.
2. Wire model_name → 🐸 Load Library by Name.
3. That node matches the stem against your Library entry names (exact first, then substring), applies the matched entry's LoRA stack to the MODEL/CLIP, and outputs its positive/negative text.
4. Stack a character prompt on top via the Library's passthrough inputs.

Switch the checkpoint dropdown, queue, and the quality prompt swaps itself. It's a small thing - one string output - but it turns the Library from a thing you click into a thing that's always correct.

The inputs and outputs

Inputs are the same four dropdowns as FrogLoader, with one tooltip worth repeating: diffusion_model's note literally says "Filename stem is output as model_name. Wire model_name → Load Library by Name to auto-load the matching Library entry." Outputs are MODEL, CLIP, VAE, and the model_name string. Note there's no separate "name" input - the name is derived from whatever checkpoint is in the dropdown, which is the whole point.

Installing it

Same pack, same install:

cd ComfyUI/custom_nodes
git clone https://github.com/RabbitThatIsPink/FrogNodePack

or via ComfyUI Manager. Restart, hard-refresh, no extra dependencies.

Where people trip

The main confusion is expecting model_name to be the display title you gave something in a manager UI. It's not - it's the file stem, exactly as it appears on disk. So if your checkpoint files live under a subfolder, expect the stem to include the relative path. That matters for matching: Library entry names should be built from what this output actually produces. Also, if you wire this into Load Library by Name and nothing seems to load, check the console - the by-name node prints a warning when no entry matches, and "nothing happened" with a console warning is the correct behavior, not a bug.