Integrations
ComfyUI custom nodes for making typed yes-or-no, categorical, or rating decisions across images, video, and text using local multimodal models.
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
nomadoor’s nodes use ComfyUI’s own engine with imajev-4b, Clef-flash and the d3 family. Local weights require suitable hardware and several gigabytes of storage. d3 can slow sharply on long videos. The separate Moment Search node checks frame-visible conditions at two frames per second, not movement direction, and can miss events shorter than half a second. These are different nodes from ComfyUI-SystemOne, and model confidence is not a guarantee.
Install through ComfyUI Manager and download a supported model as the README describes. Start with the sample that sorts photos with and without people, changing the question for your own images. For finding a person in a video, try Moment Search, but check its results: brief appearances can fall between sampled frames.