What this is
The image-to-video example that ships with the ModelRunner ComfyUI node pack. Load Image feeds a hosted image-to-video node, and Save Video writes the clip. Nothing is downloaded and no local GPU is used.
Disclosure
We make the pack. It calls a paid service: every run is billed per request to a ModelRunner balance, and you need your own API key. There is no subscription, and a failed run is not charged.
Setup
Step 1: Install "ModelRunner" from ComfyUI Manager (ComfyUI 0.12.1 or newer).
Step 2: Set MODELRUNNER_API_KEY in your environment before starting ComfyUI. The key is never a node widget, so it is not saved into workflow files or image metadata.
Step 3: Open this workflow, pick an image, and queue.
Notes
• Seeds start on fixed. Re-queueing an unchanged graph returns the cached clip and creates no new request.
• Interrupting stops the wait, not the job. A submitted run still completes and is charged.
• The pack also has typed nodes for image, audio and 3D models in the public catalog, and a "Run (any model)" node that takes an owner/alias endpoint and a JSON input.
• Inputs and results are stored on ModelRunner's media host under unguessable URLs.
Links
Source: https://github.com/modelrunner/modelrunner-comfyui
Guide: https://modelrunner.ai/comfyui
Walkthrough: https://modelrunner.ai/blog/comfyui-image-to-video-workflow

