META · SAM 3

SAM 3 Video: mask and track anything you can name

SAM 3 is Meta's newest Segment Anything model, the third generation of the SAM family and the first built for open-vocabulary video. Name objects in a prompt and it masks and tracks each one through every frame, with box and point prompts to refine any selection. Available now in Picsart's AI Playground.

Video
Name the objects to mask and track...

Videos masked with SAM 3

SAM 3 Video sample clip (placeholder creative)
SAM 3 Video sample clip (placeholder creative)
SAM 3 Video sample clip (placeholder creative)
WHAT IS SAM 3

What is SAM 3 Video?

SAM 3 is the third generation of Meta's Segment Anything models, built for images and video. Give it a short noun phrase and it detects every object matching the concept, draws a precise mask around each one, and assigns it an ID so the mask stays with that object as the video plays. Visual prompts work too - points, boxes, and mask examples - and a box around one example tells SAM 3 to find all the rest.

  • Bring your own video
  • Mask every object you name
  • Keep editing in Picsart
SAM 3 masking objects named in a prompt (placeholder creative)
ALL MATCHES, NOT ONE

Every match gets its own mask

Older segmentation tools selected one object per click. SAM 3 is exhaustive: prompt "red baseball cap" and every cap in the shot gets masked, each tracked separately. The model reads an open vocabulary - over 50 times more concepts than earlier benchmarks - so nuanced phrases like "the striped red umbrella" land on exactly the right object. When a frame needs a correction, refine the selection with a click.

Multiple matching objects masked at once (placeholder creative)
BUILT FOR MOVING FOOTAGE

Track up to 16 objects in one pass

The SAM 3.1 update made video the headline. Object multiplexing tracks up to 16 objects in a single forward pass, and global reasoning processes them together instead of one by one - twice the speed of the first release, at near real-time rates. Masks hold their identity through motion, occlusion, and re-entry, which is what makes per-object work possible: isolate a subject for color, blur everyone else, or swap what a tracked object looks like.

Objects tracked through a moving shot (placeholder creative)
INSIDE AI PLAYGROUND

SAM 3 Video in Picsart AI Playground

SAM 3 Video lives in Picsart's AI Playground alongside 150+ other AI models - upload a clip, name what you want selected, and build on the tracked masks right next to your other creative tools, with no third-party API keys or separate subscriptions. Reach it whichever way you work: on the web, in the desktop app, and via CLI, MCP, REST API, and SDK, so a selection you prototype here drops straight into an automated pipeline.

SAM 3 inside Picsart AI Playground (placeholder creative)

What you can do with SAM 3 Video

Three ways to put promptable masking to work.
Effect applied to a masked subject only (placeholder creative)
  • Mask the person or product once and apply color, glow, or blur to just that layer while everything else stays untouched.

How to mask and track objects with SAM 3

Upload a clip, name the objects, get a tracked mask for every match, and finish in Picsart.

How to

Upload your video

Open SAM 3 in Picsart AI Playground and add the clip you want to work on.

Step 1: Uploading a video to SAM 3 (placeholder creative)

SAM 3 Video FAQ

Answers about masking and tracking objects with SAM 3 in Picsart.

  • SAM 3 is Meta's Segment Anything Model 3 - a segmentation model that detects, masks, and tracks objects in images and video from text or visual prompts. The SAM 3.1 update, released in March 2026, doubled its video tracking speed.

Pricing

Try a few clips or build a repeatable masking workflow for edits, effects, and campaigns.

Pro

Most popular

AI tools for everyday creative work.

$15 $10.5/mo
Billed yearly
You save $54 with yearly
  • Access to all photo & video editing features
  • Advanced background & object removal
  • Parallel video generations with the world's most powerful AI video models
  • Unlimited image generations with Flex.2 Klein
  • 1-tap image enhancer
  • Millions of stock photos & Getty video clips
  • Selection of trendy fonts, text styles & stickers
  • Thousands of premium templates
  • Support for 3+ brand kits
  • Bulk edit up to 50 images at once
  • 100 GB of cloud storage
New features:
  • Auto-generate content from your terminal or agent with the Picsart CLI
  • Use Picsart inside Claude Code, Cursor, and ChatGPT via MCP — coming soon
  • AI agents for multi-step workflows and batch generation — coming soon

Ultra

Most powerful

Heavy AI usage for creators & teams.

$45 $24.5/mo
Billed yearly, per seat
You save $246 with yearly
  • Everything in Pro
  • Early access to advanced AI features
  • Leading AI models to design & automate workflows (Nano Banana, Veo 3, Seedance 2.0 & more)
  • Parallel video generations with the world's most powerful AI video models
  • Unlimited image generations with Flex.2 Klein
  • Support for 10+ brand kits
  • Add team seats
  • Create ad variations and localize
  • Track ads performance
  • 2000 credits for API services
  • Bulk edit up to 100 images at once
  • 300 GB of cloud storage per seat
New features:
  • Auto-generate content from your terminal or agent with the Picsart CLI
  • Use Picsart inside Claude Code, Cursor, and ChatGPT via MCP — coming soon
  • AI agents for multi-step workflows and batch generation — coming soon