comfyui_segment_anything

Why we included this project

Anyone building image editing or generation pipelines in ComfyUI can cut objects out of a photo using nothing more than a text description. The node combines GroundingDINO, which locates the described object in the frame, and SAM, which produces a tight mask around it, then exposes the mask as a workflow step that feeds the rest of the graph. It is a direct port of sd-webui-segment-anything, and output matches that extension, so people switching between the two ecosystems get consistent results. Several SAM variants are supported, including the lightweight MobileSAM, which keeps masks fast on modest hardware. For teams that would rather drive selection with words than draw masks by hand inside a node-based toolchain, this is a clean drop-in.