Available now

SAM 3.1 Image: object segmentation

Meta's SAM 3.1 finds the objects you name or point at in an image and cuts each one out as a mask.

Model ID

Input

Fields marked * are required. Review the price before generating.

The image to be segmented

Uploading a file needs an account. Create account

JPEG, PNG, WebP · up to 25 MB

Text prompt for segmentation

Example prompt. Edit it or write your own.

Additional settings

List of point prompts

Whether to include mask confidence scores.

Whether to include bounding boxes for each mask (when available).

Estimated total$0.02

You sign up first. Your settings are kept.

About this model

Overview

Upload an image and say what to select, in a few words such as "wheel" or "yellow taxi". It finds every match. You can also mark objects with points (on the object or on the background) or with boxes, and give points and boxes the same object number to refine one object.

By default the mask is applied to the image. You can also ask for several masks, a confidence score per mask and a box around each one.

What you can build

  • Masks for editing one object in a photo
  • Cut-outs of every product on a shelf at once
  • Selections to feed into inpainting or relighting
  • Labeled regions for an image dataset

Prompting tips

  • Name the object plainly; add a color or position when there are several kinds.
  • When the text alone picks the wrong thing, add a point on the object to steer it.
  • Turn off apply mask when you want the black and white mask itself.

What to check

Inspect thin edges, holes and nearby objects included in the mask.

Inputs

  • Image (required): URL of the image to be segmented.
  • Prompt (optional): Text prompt for segmentation.
  • Point Prompts (optional): List of point prompts.
  • Box Prompts (optional)
  • Apply Mask (optional): Apply the mask on the image.
  • Output Format (optional): The format of the generated image. Options: jpeg, png, webp.
  • Return Multiple Masks (optional)
  • Include Scores (optional): Whether to include mask confidence scores.
  • Include Boxes (optional): Whether to include bounding boxes for each mask (when available).

Output and limits

  • File formats: JPEG, PNG, WEBP.
  • Small or dense text in the image can come out misspelled.
  • Hands, crowds and complex poses can show artifacts; generate a few variations and pick the best.
  • Some prompts or outputs can be refused by safety checks.

Price guide

  • 1 request: $0.02 / request