Nodes/ComfyUI_Mira/Checkpoint Loader with Name
ComfyUI Node

Checkpoint Loader with Name

Checkpoint Loader with Name (CheckpointLoaderSimpleMira, ComfyUI_Mira)

By mirabarukaso·Created 2 years ago·Updated 5 days ago· 206
Checkpoint Loader with Name
    • MODEL
    • CLIP
    • VAE
    • model_name
    ckpt_name

    This is ComfyUI's standard Load Checkpoint node with one extra wire coming out of it: the checkpoint's name, as a string. That sounds trivial until you've tried to figure out which base model made an image you generated three weeks ago. Checkpoint Loader with Name exists so the model name can travel downstream into your metadata and image saver, instead of living only in the node widget where nothing can read it.

    It's a Util node from ComfyUI_Mira, and it fits the author's recurring theme: mirabarukaso builds a lot around getting model and LoRA names into the saved file so sites like CivitAI can identify what produced an image. The pack literally includes a name-stacking LoRA loader for the same reason ("any website that could not identify all LoRAs' information from image... only happens with ComfyUI"). This checkpoint loader is that idea applied to the base model.

    How it works

    Functionally it loads a diffusion checkpoint exactly like the core node - same MODEL, CLIP, and VAE outputs, same denoising role in the graph. The addition is a fourth output, model_name, that emits the selected checkpoint's filename as a STRING. You route that string into an image-saver node that embeds generation metadata, and now the output PNG records which checkpoint made it. Nothing about the actual model loading is different; you're just exposing a piece of information the standard node keeps to itself.

    Inputs and outputs that matter

    • ckpt_name - the checkpoint to load, picked from your models/checkpoints folder. Same dropdown you already know.

    Outputs:

    • MODEL / CLIP / VAE - the usual three; wire them exactly as you would from Load Checkpoint.
    • model_name - the new one. A STRING of the checkpoint name, meant for your saver's metadata or a ShowText node.

    That's the whole delta. If you don't care about capturing the model name, the core loader does the same job; the moment you want the checkpoint recorded in your output, this saves you a manual PrimitiveNode-and-type dance.

    Installing it

    ComfyUI Manager: search ComfyUI_Mira in the Custom Nodes Manager, Install, restart. Manual: cd ComfyUI/custom_nodes && git clone https://github.com/mirabarukaso/ComfyUI_Mira.git, then restart ComfyUI. If it errors on load, pip install -r requirements.txt in the pack directory and restart. The node itself pulls no models - it just loads the checkpoints you already have.

    Common issues

    The dropdown is empty until you actually have checkpoints in ComfyUI/models/checkpoints and ComfyUI has scanned them; if you just dropped a file in, restart so it's picked up. If the node doesn't appear at all after install, do a full restart rather than a browser refresh, and check the console for a dependency error that a requirements.txt install would fix.

    The subtler thing to get right is the reason you'd use it: the model_name string only helps if you actually connect it to a saver that writes metadata into the file. Wired to nothing, it's the same as the plain loader. And note it emits the raw filename - if your checkpoints have cryptic names, the metadata will be cryptic too, so this is a nudge to keep sane filenames. Pair it with Mira's name-stacking LoRA loader if you want both the base model and every LoRA captured in the same output.

    CategoryMira/Util

    Inputs (1)

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

    Outputs (4)

    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.
    model_nameSTRINGThe model_name used for Image Save node to save model name.