ComfyUI-YoloWorld-EfficientSAM

Why we included this project

This node pack tackles a common ComfyUI gap: finding and isolating specific objects in an image or video. It pairs YOLO-World's open-vocabulary detection with EfficientSAM's segmentation, so you type the categories you care about and get bounding boxes plus per-object masks back, no fixed class list required. The V2.0 nodes added the practical controls: merge masks or keep them separate, extract a single mask by index, tune confidence and IoU thresholds, and toggle class-agnostic NMS. Video input works too, and the Detector Provider node integrates with Impact-Pack for more involved setups. If you already run ComfyUI for generation or editing, this is a low-friction way to add grounding and masking steps without leaving your graph.