Runware Import Model

Most of this pack is "pick a model from the catalog, run it." Import Model is the part that answers "but what about my model?" It's a builder that describes a model you want available on the Runware platform - your private fine-tune, a community checkpoint you use constantly - so the custom-model nodes can actually run it. Nothing generates here; you're registering a model. The output is a RUNWARE_IMPORTMODEL value that you wire into the importModel socket of a custom checkpoint node (the Runware/Custom models nodes, one per architecture).

The inputs

• air - required STRING. The tooltip spells out the format: provider:model@version. This is the address of the resource on the platform. If you've seen AIR identifiers elsewhere in this pack - bria:20@1, runware:110@1 - this is the same scheme, applied to your model.
• name - required STRING. Human-readable name for the model.
• private - default true. Whether the model is private to your account. You almost certainly want it private; the default is already the safe choice, which is a nice change of pace.
• version - STRING, optional. Version string for the model.
• uniqueIdentifier - STRING, optional. A unique ID for the model, when you need to disambiguate.
• heroImageURL, shortDescription - cosmetic metadata for your model's page/listing. Optional, safe to skip.

Where it fits

The custom-checkpoint path exists because "every Runware model as a node" only covers what Runware curates. If you live on a specific community fine-tune - an Illustrious derivative, a Pony mix, your own trained LoRA folded into a checkpoint - the curated list is the wrong list, and this builder plus a Runware/Custom models node is the door in. Think of air as the foreign key: the custom node that actually runs inference looks up this AIR, and the import step is what makes that lookup succeed.

Output and wiring

Output is RUNWARE_IMPORTMODEL. It feeds a custom checkpoint node's importModel socket - it does not feed a standard catalog model node, because those already know their AIR.

Install

Same pack install as everything else:

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

Restart, add your Runware API key (Settings → Runware API key, or RUNWARE_API_KEY).

The honest take

Most people will never touch this node, and that's fine - if you're happy with the curated catalog, you don't need it. It's also the node where the "cloud inference" framing gets real: the whole point is that your checkpoint stops being a file you maintain and becomes a resource you can call from any machine with a key. The friction is that the AIR format (provider:model@version) is exact and unforgiving, and the import process itself - getting the weights onto the platform in the first place - lives outside this node, in Runware's dashboard/docs. This node is the "declare it exists" step, not the "upload it" step, and knowing that up front saves you from wondering why the builder alone didn't make your model runnable.