ComfyUI+SegmentAnything2: Achieve Video Cutout Freedom! Multi-Object Support and Occlusion Handling!

Meta, formerly known as Facebook, launched the Segment Anything model last April, which has proven to be very useful. I have used it in multiple ComfyUI workflows to achieve precise "cutouts". Now, more than a year later, the upgraded Segment Anything 2 is here! It not only shows significant improvements in image segmentation speed and effects (performance nearly tripled), but it also expands from the image domain to the video domain. By storing and utilizing previously processed video frame information, it provides real-time and accurate object segmentation for dynamic video content.
In this video, I will guide you step-by-step from downloading and installing the model and plugins to completing workflows for marking objects in single images, marking objects in videos, and marking multiple objects simultaneously in videos!
Custom Node: github.com/kij...
Workflow: drive.google.c...
Model: huggingface.co...