AIHub Meta Set Exported Model Image

AIHubMetaSetExportedModelImage is the model-catalog sibling of the pack's other "set the image" nodes: it stamps a PNG thumbnail onto an already-exported model in the otavanopisto ComfyUI-aihub-workflow-exposer system. The AIHubMetaExportModel node (the model equivalent of AIHubMetaExportLora) registers a checkpoint or diffusion model with the client-facing catalog - name, family, defaults, VAE, the works - and optionally writes an image at the same time. This node is the "set the image only" step, so you can fix the model's cover art without redoing the metadata.

In this pack's world, a model's catalog image is what clients show when the user browses "which model do I want to generate with." If the model was exported without one, the user picks blind. The README is explicit that this standalone node exists for exactly that job: "Equally to the lora, an image can be chosen for the model, and the standalone node AIHubMetaSetExportedModelImage can be used for that."

How it works

Two required inputs, no outputs:

• model - the id of an already-exported model, from a dropdown populated by ComfyUI/aihub/models. You can only set an image for a model that's in the catalog.
• image - any IMAGE. First frame, converted to PNG, saved as {id}.png in the model's aihub folder. No image → the source raises "No image specified."

That's the whole mechanism. Run it, the file overwrites, done.

Why you'd reach for it

Because the export nodes are all-or-nothing at export time, this is the iterate-in-place tool. A workflow that exports a model with a placeholder thumbnail can be republished with a proper example image any number of times. It's also useful when you want a consistent cover style across a model family - generate one set of thumbnails in a batch workflow, then feed them to this node one by one. The model's metadata JSON is untouched, so you never risk corrupting the default_cfg, default_steps, or family fields you carefully configured at export.

Gotchas

The dropdown reads from the exported folder, so a model you haven't exported yet won't appear - run AIHubMetaExportModel first or check ComfyUI/aihub/models on disk. And remember this pack's catalog model: exporting is only necessary when a workflow should accept any model in a family or group. If your workflow is pinned to one model, you can load it directly with AIHubUtilsLoadModel and this whole catalog path - image included - is optional.

Install is the pack standard, no requirements.txt, nothing to download:

cd ComfyUI/custom_nodes
git clone https://github.com/otavanopisto/ComfyUI-aihub-workflow-exposer

Restart ComfyUI. It's a two-input, one-file utility - nothing glamorous, but if you're maintaining a client-facing model catalog it's the difference between shipping a model with a real preview and shipping one with a blank card.