Nodes/ComfyUI-IcyHider/Icy CheckpointLoaderSimple
ComfyUI Node

Icy CheckpointLoaderSimple

The one loader every workflow starts with — CheckpointLoaderSimple, minus the mystery

By icekiub-ai·Created 9 months ago·Updated 9 months ago· 31
Icy CheckpointLoaderSimple
    • MODEL
    • CLIP
    • VAE
    ckpt_name

    IcyCheckpointLoaderSimple is the IcyHider clone of the node you'll see at the top of almost every workflow you ever download: CheckpointLoaderSimple. If you're new, this is the "start here" loader - one dropdown, three outputs, and everything else in the graph hangs off it.

    What it does

    One field:

    • ckpt_name - the checkpoint file, from a dropdown populated by models/checkpoints. The tooltip says it plainly: "the name of the checkpoint (model) to load." Drop a .safetensors (or .ckpt) into that folder, restart, and it appears in the list.

    Three outputs, and each one goes somewhere specific:

    • MODEL - "the model used for denoising latents." This is the heart of the graph; it feeds your sampler and any model modifiers.
    • CLIP - "the CLIP model used for encoding text prompts." It drives your CLIPTextEncode nodes. The tooltips in this schema are the author's own, and they're accurate: this is the text-understanding half.
    • VAE - "the VAE model used for encoding and decoding images to and from latent space." It converts your finished latents back into viewable pixels (and encodes images for img2img).

    That's the entire trick of a checkpoint: one file that carries the denoiser, the text encoder, and the VAE, and this node unpacks all three at once. It's the "simple" because there's no config dropdown, no architecture guessing - the loader inspects the file and figures out what it is.

    How to use it

    Wire MODEL into a KSampler (optionally through LoRA or other model patches), CLIP into one or two CLIPTextEncode nodes (positive and negative), and VAE into the VAEDecode that turns your latent into an image. If you're following a beginner tutorial, this is the node at the top left.

    Where people get burned: the empty dropdown. Every time, it's one of two things - the file isn't in models/checkpoints, or you installed/added the file after ComfyUI started and haven't restarted. Also: newer model releases increasingly ship without the "simple" convenience (separate UNet, separate text encoder, separate VAE). If your model isn't a single checkpoint file, CheckpointLoaderSimple can't load it - that's what UNETLoader + CLIPLoader + VAELoader are for.

    The "Icy" caveat

    Outputs are MODEL/CLIP/VAE - no image preview for IcyHider's hider to hide, so the Icy wrapper is pure relabeling. The pack clones every core node at startup; this one is just along for the ride. Functionally it's byte-for-byte the core node.

    Install

    ComfyUI Manager → search "IcyHider" → install → restart, or:

    cd ComfyUI/custom_nodes
    git clone https://github.com/icekiub-ai/ComfyUI-IcyHider.git
    

    Restart ComfyUI. No Python dependencies and no model bundled - the checkpoint is your file. The node is under IcyHider Comfy Core. Empty dropdown, blank screen, one restart later: solved.

    CategoryIcyHider Comfy Core

    Inputs (1)

    NameTypeDefaultDescription
    ckpt_nameCOMBOThe name of the checkpoint (model) to load.

    Outputs (3)

    NameTypeDescription
    MODELMODELThe model used for denoising latents.
    CLIPCLIPThe CLIP model used for encoding text prompts.
    VAEVAEThe VAE model used for encoding and decoding images to and from latent space.