Nodes/ComfyUI-TechsToolz/🖥️🛠TechsToolz: Model Name Extractor
ComfyUI Node

🖥️🛠TechsToolz: Model Name Extractor

Which checkpoint actually made this image? One wire, zero typing

By robin-collins·Created about a year ago·Updated about a year ago· 0
🖥️🛠TechsToolz: Model Name Extractor
  • model
  • model_name

Every ComfyUI user has stared at a finished image and asked the boring but important question: what model made this? The answer is in your graph, but it's buried in the loader node, and once the workflow is shared or you've rebuilt it a dozen times, "which checkpoint" stops being obvious. This node answers it directly: feed it a MODEL object and it hands you the model's filename as a plain string.

It's the second of two nodes in ComfyUI-TechsToolz, Robin Collins' small personal pack (version 0.0.2, "a collection of what I wanted!"). It's a genuinely tiny utility - one input, one output - and that's the whole point. It's not a node you notice; it's a node that makes the other node work.

How it works

ComfyUI model objects carry their name in different places depending on how they were loaded, and the extractor checks them in order until one hits:

  1. gguf_filename - set by GGUF loaders (the city96 ComfyUI-GGUF pack), so quantized Flux and video models resolve too.
  2. The model_config.unet_config.model_path buried inside a standard CheckpointLoaderSimple result.
  3. A direct model_path or filename attribute, as fallback for other loader types.

If nothing matches, you get "unknown_model" rather than an error. It strips the directory and extension where it can, so what comes out is a clean name like juggernautXL_v9 or flux1-dev-Q8_0, not a full path.

What it wires into

The obvious home for the output is the same pack's Save Image w/Metadata node, which needs the model's exact filename to build its Model hash: metadata - and crashes if you type it wrong. Feed the extractor's string into its modelname input and the whole thing stays correct even when you swap checkpoints. A second string output could also go to any text display node, or into a filename pattern with %model.

If you're only running the built-in checkpoint loader and never touch GGUF, this node is arguably doing the work for you that you could do by hand. The moment you juggle quantized and full-precision checkpoints in the same workflow, though, it earns its keep - the gguf_filename path is exactly the one most naive "model name" nodes miss.

Installing

Same pack as its sibling, same install:

cd ComfyUI/custom_nodes
git clone https://github.com/robin-collins/ComfyUI-TechsToolz.git
cd ComfyUI-TechsToolz
python install.py

Or search "ComfyUI-TechsToolz" in ComfyUI Manager and restart. Dependencies are just piexif, numpy, and Pillow - nothing heavy, no model downloads.

Caveats

The output isn't always a bare filename: for GGUF models it returns the gguf_filename attribute as-is, so you may see a .gguf filename with its extension depending on how the loader set it. And this pack is young with a tiny install base - zero community discussion so far - so if the extractor returns unknown_model, it's not a bug report you'll find help for. Check your loader type against the attribute list above; on the standard loaders it's boring and reliable, which is exactly what you want from a utility node.

Category🖥️🛠TechsToolz

Inputs (1)

NameTypeDefaultDescription
modelMODEL—

Outputs (1)

NameTypeDescription
model_nameSTRING—