AI Smart Image Segmentation

AI Smart Image Segmentation is a Figma plugin that uses the Segment Anything Model together with Transformers.js to cut objects out of images. You click on the picture to add positive points for what you want to keep and negative points for what you want removed, and the segmentation updates in real time as you go. Multiple mask mode handles complicated images where one selection is not enough. The result is exported straight to the Figma canvas with customizable outlines. The plugin is free to use and open source.

The click-based interaction is what makes it faster than a pen tool for rough cutouts. Instead of tracing an edge point by point, you indicate roughly what belongs to the subject and let the model work out the boundary, then refine with a few negative points where it has gone too far. For a product on a plain background this is close to instant, and for a busy scene it still saves most of the manual work.

- Segment an image by clicking rather than drawing a path

- Refine the selection with positive and negative points

- Handle complex subjects using multiple masks

- Export the cutout into Figma and keep designing

- Work on the image without leaving the design file

- Designers pulling elements out of photos for layouts

- Artists building compositions from existing imagery

- Advertisers preparing product and model cutouts quickly

- Social media creators making quick graphics

Fine edges such as hair, glass, or soft shadows still need manual cleanup, because a click-based mask is an approximation and the model can bleed into the background. It runs inside Figma, so it is not a standalone image editor and it will not help outside that environment. The plugin depends on upstream model code and the Figma plugin platform, so updates to either can change how it behaves. Because it is open source, check the repository if you need to understand what data the plugin processes before you use it on client work.