Ino Get Image Model Download Config

Downloading a checkpoint is usually the most manual part of standing up a new workflow: find the repo or Civitai page, match the filename, drop it in the right diffusion_models subfolder. Ino Get Image Model Download Config compresses all of that into a dropdown - pick a base model by name, get back a JSON config string that the pack's downloader can act on without you touching a browser.

The output, ModelConfig (STRING), is designed to feed InoHandleDownloadModel, which reads the config, resolves the host (HuggingFace, S3, or Civitai), pulls the file into the right model directory, and returns the path. From there you can wire it into a loader node and keep the whole bootstrap inside the graph. It's the image-model sibling of the pack's LoRA, VAE, CLIP, video, and ControlNet download-config nodes - same pattern, different CSV.

How it works

The dropdown is populated from data/files/image_models_files.csv, shipped with the pack. Select a row's name and the node serializes that row to JSON. The CSV columns reveal the config shape: name, weight_type, host, model_type, model_subfolder, repo_id, filename, subfolder, repo_type, revision. Same schema the other config nodes emit, so one downloader handles them all.

Inputs and outputs

• enabled - pack-wide toggle; off returns an empty string.
• model - the dropdown of image models from the bundled CSV.

Output: ModelConfig - JSON for the download chain.

Installing it

Part of ComfyUI-InoNodes (Inoland). ComfyUI Manager → search "ComfyUI Ino Nodes" → install → restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/nobandegani/ComfyUI-InoNodes
cd comfyui_ino_nodes
pip install -r requirements.txt

Restart after. V3-schema pack, so keep ComfyUI current (v0.18.1+). inopyutils auto-installs. If the model you pick hosts on Civitai and is gated, set CIVITAI_TOKEN before the download step runs.

Common issues

Same two traps as the other config nodes. First, this node doesn't download anything - no wire out means no file, just a JSON string sitting on the canvas. Second, the dropdown is only as long as the shipped CSV, and it's curated, not exhaustive. The bundled list is Flux-leaning - it ships entries like Flux1dev from Black Forest Labs plus a handful of Flux 1 dev fine-tunes hosted on Civitai - so don't expect every checkpoint on the internet to be there. No "refresh" exists; the list updates with the pack.

One practical note: image models are multi-gigabyte downloads, so if you're automating several workflows with this, be ready for the first run of each to stall while it pulls the file. That's the downloader doing its job, not the node hanging. And when the config's model_subfolder says flux1dev, that's where the file lands under your diffusion_models folder - keep that path in mind when you're writing custom loader nodes.