SP_Name_Unet

Older SD 1.5/SDXL workflows load a single checkpoint file that bundles the diffusion model, CLIP, and VAE together. Newer architectures - Flux especially - increasingly ship those pieces as separate files instead, loaded through ComfyUI's dedicated diffusion-model loader rather than a single Checkpoint Loader. SP_Name_Unet is for that split-file world: it picks a filename from wherever those standalone diffusion-model weights live and returns it as text, no loading involved.

Why you'd want just the name

Once model, CLIP, and VAE are three separate files instead of one, keeping track of exactly which combination produced a given output gets harder, not easier - three independent choices instead of one. This node exists to make one leg of that combination legible: tag output filenames or logs with which diffusion-model weight was actually loaded, or drive a switch node off that choice, without paying to load a multi-gigabyte file just to confirm its own name.

Inputs and outputs that matter

• name - a dropdown populated from your diffusion-model weights folder.
• Output: STRING - the selected filename.

How to install it

Through ComfyUI Manager: search "SP-Nodes," install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/bananasss00/ComfyUI-SP-Nodes

Restart ComfyUI once it's cloned. No extra dependency for this node itself.

Common issues & troubleshooting

Dropdown is empty even though you have a checkpoint installed. This node reads the folder for standalone diffusion-model weights, which is a separate location from your regular checkpoints folder - a single-file checkpoint (SD 1.5, SDXL, most all-in-one packages) won't show up here at all, because it isn't split out. This node only has something to show once you've installed a model distributed as a separate UNet/diffusion-model file.

Not sure if your model needs this split-file approach at all. That depends entirely on how the checkpoint you downloaded is packaged - check the model's own download page or model card. If it's a single .safetensors bundling everything, use a normal Checkpoint Loader and skip this node entirely; it's only relevant once model, CLIP, and VAE genuinely ship as separate files.

Confusing this for the actual model loader. This node stops at a filename - it never loads weights into VRAM and never produces a MODEL output. The real diffusion-model loading still needs ComfyUI's own dedicated loader node wired into your sampler; this one is for labeling and logic elsewhere in the graph, not for the pipeline itself.

Multiple similarly-named files and you're not sure which is which. Split-file distributions often ship several precision/quantization variants of the same model side by side - fp16, fp8, a GGUF quant - and the filenames don't always make the difference obvious at a glance. If VRAM or speed is the concern, check the file size in your file browser before picking one here; this node's dropdown shows names only, not sizes.